Return-to-zero shift register FIRDAC

[Hans Polak]

FFT’s as such, especially when in the ppm range are meaningless, but the point in this case were the unexplainable not directly THD related anomalies in the spectrum that could be caused by either the digital or the analog section.
That’s where the FFT could reveal something.
Whether this is perceivable, who knows.
But its interesting to find the reason.

Hans

 

[Tfive]

How about all the brilliant minds posting recently - put together your strength of force and try to come up with some more things we can actually measure, an addition to the FFT measurements? I don't think that's strictly impossible, just - probably - hard.

Why are new measurement procedures kind of important? not all new and upcoming builders have time and ears to qualify their builds. Take me, for some part, into this camp. If you, the experienced developers of electronic circuits, could team up and shed some light on how modulator noise in conjunction with non ideal analog circuitry after the DAC could be quanitified - a lot of gain for the whole communitiy could arise.

I know lots of proprietary technology is involved, driven by commercial purposes and NDAs. Please try to share with us - as far as possible. Minds like me will take it all up. And in the end, if we - very therotically - would create a better DAC that every one can build easily - isn't that the whole purpose of this forum, provide the best sound possible to anyone who wants wo achieve it?

 

[Hans Polak]

Well, that’s excactly what made me construct a bridge between DSD file and LTSpice, which will enable me to play a real .dsf files over the Firdac and Filter module in LTSpice.
This may or may not shed some light on things like aliasing or whatever could be causing anomalies.
In LTSpice it’s much easier to investigate these things and to compare results with the measured recordings while using the same .dsf files.
I did the first run today and the model is working properly
Will comeback when results are ready to show.
So instead of talking about possible problems this is hands on.

Hans

 

[Markw4]

@Tfive
Well, for starters there is this claim:

A new distortion measurement technique which uses a signal more closely resembling program material was presented. It is considerably faster than other distortion measurement techniques and allows simultaneous measurement of frequency response, phase, crosstalk, and noise.

Which is described in the attached.

Not to say that's everything...

If we had a standard way of evaluating spectral line widths and or noise floor modulations, that might be a step in the right direction. Research to correlate those measurements with human perception is not so easy however.

More posting of distortion residuals might help too.

All that having been said, and at this point for the individual designer, Thor already talked about "going to see the elephant."
https://en.wikipedia.org/wiki/Seeing_the_elephant

According to @soundbloke "we all have a capability to train our hearing on minute details that an untrained listener would find inaudible."

I have offered to accept visitors here. Bring a dac and we will compare it to the one here and to vinyl. You will probably notice some details you didn't hear before, but maybe not. It varies from person to person in terms of what soundbloke speaks of.

EDIT: Also, presumably it should be possible to measure the phase distribution of FFT frequency content associated with noise floor modulation and or spectral line widening. The phase distribution of noise may be very telling if one considers the possibility of non-gaussian bell curves for some frequency distributions. White noise may not be smooth, actual hearing of small differences may not as unusual as expected, etc. Maybe some signal correlated noise can be more audible than might otherwise be expected due to, say, some crest factor effect. If so, doesn't mean it would look any different from smooth noise on an FFT that discards phase information.

 

[Tfive]

All that having been said, and at this point for the individual designer, Thor already talked about "going to see the elephant."
https://en.wikipedia.org/wiki/Seeing_the_elephant

Quoting wikipedia here, as this phrase was not previously known to me:
"The phrase seeing the elephant is an Americanism which refers to gaining experience of the world at a significant cost."

If some of us took significant the time and effort to "see the elephant" it would be most appreciable if the outcomes of this effort and time could be shared with the commuinity at large. Why? I have several friends that share a common dedication to good music and proper reproduction of it. Several times these friends have come to my place, not knowing that anything had changed (maybe improved) in my system and ALL the times people noticed. I intentionally never gave them a clou beforehand. So, why exclude these people from insights gained by our community? These people just want to listen to good music reproduction and I bet you they will neither be able to understand any technical stuff we talk about nor will they - given their monetary background - ever be able to purchase, or not even listen to such high end and pricey gear that is currently considered "top of the line". We're speaking of DACs that sometimes cost way upwards of 1000€, sometimes ten times of that or more.

Are we intentionally trying to build up an elite of people who will ever be able to listen to such systems - or - in the true DIY spirit - are we trying to build proper building blocks that could be manufactured in a DIY fashion to enable these people to have an ability to better enjoy music at an affordable level? Maybe built by less experienced people like I am, could be providing them with working, enyoiable circuitry at lower cost than commercial solutions.

 

[Markw4]

...why exclude these people from insights gained by our community?

I would recommend not to exclude them. Rather train and qualify them into your own skilled listening panel.

Its because of this: Nobody can always trust themselves. You need people to check on you, to keep you honest. And when I say you, I mean me too. Its a valuable thing to have and worth the trouble to put together.

EDIT: Its simply a matter of recognizing we are still at a state of knowledge where biological sensors can be better to have than not have. We don't know how to do all the same computations in a computer.

 

[Tfive]

But you cannot / dont't want to make the listening equipment available to them because they did not "see the elephant" themselves?

 

[Markw4]

No. Why would you think that?

Some people hear small details more naturally. Some other folks have to spend time learning one small detail at a time. It just varies.

People who serve on a panel should maybe get something for their efforts. Depends how much they put into it. Also depends what the best you can do for then is. Can you give them one of your designs for free for participating? I don't know. Can you give them a board and case and let them build their own? These are details you have to work out for your individual situation.

 

[ThorstenL]

Yeah. But a lot of people find it very hard to believe because they feel if such were true it would be well explained by Self and or Cordell, the authors of the bibles of audio.

BUT IT IS IN THERE!

It's in Mr. Self's work. All that is needed is actually reading the whole lot, average reading comprehension for a 5th grader and the ability to deduce the implications.

If for example Mr. Self states that final slope of the rise of distortion for a "blameless amplifier" with frequency past the knee for a given circuit is something like a 8th to 12th order function, the systemic implications should be blatantly obvious.

And yes, in a correctly designed and bandwidth limited analogue linear amplifier it is possible to avoid to actualising the resulting risks.

DAC I/U conversion for Multibit DAC's? Less so. DS DAC's? Are you kidding, the output of the modulator contains more noise at high frequencies past the distortion knee than the full scale signal of the DAC and the noise commonly is non-gaussian but signal correlated.

Probably the reason they don't get it is because HD FFT spurs are easy to interpret. There is also some published research on how such spurs affect perceptual SQ.

Is there? I'd be interested to read that (presuming we are not talking about Olson & DEL Shorter or Earl Geddes on HD... Even the newest of this research is decades old and merely restates the classic distortion masking that goes back nearly a century.

OTOH, a shift in the noise floor and or a shift in spectral line noise skirt width looks small on an FFT. And there is no published research on human sensitivity to such effects. Thus many people feel it is justified to assume such effects are completely negligible.

I remember when "Transient Intermodualtion" (TIM) and "Slewrate Induced Distortion" (SID) and "Phasemodulation Induced Distortion"(PIM) and "Phase Noise Induced Distortion" (PNID aka Jitter) were consider audiophile myth.

Now we recognise all of them as being audible fidelity impairments with audibility thresholds (that are under debate) nearly 5 decades after some people first suggested their eXistenZ, though we still seem to have major problems actually nailing down what the exact audibility limit is, something which parallels older known fidelity impairments such as harmonic distortion of frequency response deviations.

And so we have the eternal conflict between the myopic meter reader and 'scope jockey with the idea that because something is measures it matters and minimising it at all cost and the little kid listening and noting that the new emperor is naked (aka this witchet sounds NDFG).

And then people go right back to talking about THD and or THD+N numbers as though that's all there is a that matters.

Reminds me of the old story of Nasruddin looking for his housekeys under the streetlight.

Changing the subject a bit, referring to noise intermodulated with an audio signal as IMD may confuse people who are trained to think of IMD as exclusively the product of a signal passing through an unchanging curved transfer function. In that view HD and IME are exactly the same measurement of nonlinearity, and both are exclusively PSS.

Well, we have all sorts of ridiculous theories. There is another one.

Intermodulation is simple.

If two signals with different phase and/or frequency pass through a nonlinear transfer function (and there is no electronic device with a linear transfer function) difference and multiple products will be created depending on the two frequencies involved.

All it requires is two signals and a nonlinear transfer function. And if instead of the second signal we use noise, it's still IMD.

Maybe its because people learn simplified models first, then come to believe the models are the reality.

We see this a lot. Climate Science. Social Science. Psychology/Psychiatry, Economics.

People forget that their models are INVARIABLY gross simplifications of reality and that congruence between simplified models and reality is limited and often poor.

Once this is sufficiently forgotten, we end up not with science and engineering, but Cargo Cult Science and Cargo Cult Engineering (If anyone is interested in the origin of the terms - which I seriously doubt - read here: Some remarks on science, pseudoscience, and learning how to not fool yourself. Caltech’s 1974 commencement address" by noted Latin Percussionist and recording artist Richard P. Feynman).

At that point unlearning what is wrong about that becomes very difficult.

Learning is hard. Unlearning is harder. If unlearning strays into an area of what I call "lack of moral affordability" it becomes impossible.

“Why, sometimes I've believed as many as six impossible things before breakfast.”

Charles Lutwidge Dodgson

Thor

 

[Tfive]

I release all my designs as open source/open hardware.

People who serve on a panel should maybe get something for their efforts. Depends how much they put into it.

No, just no.

They don't have to put any effort into my "listening panel" to be eligible to receive a working circuit by me. They just request it and I will build it for them. Simply they are more than my "listening panel", they are my friends that are as much interested in good music and it's reproduction as me.

I built one of my first long time girlfriends a PCM2702 DAC way back then when it was some of the "hottest new ****" in usb dacs. So she could listen to more enjoyable music. And IMO the PCM27XX DACs were very enyoiable, otherwise I would not have recommended them.

You simply don't get it. Not everything you give has to come from something you get back. I feel very sorry for you! Over and out for now regarding your opinion.

Just tried to get some community effort together... At least YOU are not interested in that. Noted. Maybe someone else besides Hans will have a different opinion? Go for it, looking forward to that.

 

[ThorstenL]

Dear Hans,

As a matter of fact, I think that LTSpice is quite accurate in it's prediction of frequency responses.

So is pen and paper.

your previous component values were obviously not as spot on as mentioned.

Did I claim they were spot on?

When designing filters we have a range of parameters we can control.

But, I omitted this from my original post, perhaps I should not have, that what I had considered "Marcel's Holy Filter, which shall not infringed upon according to HP" was a derivative work and changed the FR.

As a result I went back and wasted my time, to go back to the original schematics and then spend as little time as possible (around 15min) to get a reasonable curve match.

The very fact that my graphs showed identical gain with Marcel's design, but the one you showed had 6dB more gain should have alerted you to the fact that that I adjusted the circuit. As might have the words:

"next comparison, circuit "Marcel original" vs. persenickity and pointless simulator finetuned ultra simple OPA1632."

(meaning I made changes)

Showing error images without mentioning the supposed background why you showed them

I am not teaching EE101, sorry.

I show what I show and I expect that those who understand electronics understand what I show and those who don't who do not, do not care enough to make themselves knowledgable.

is just as incomplete as showing Firdac diagrams without mentioning the source

So now you claim I plagiarise? Example?

showing circuit diagrams in a DiyAudio forum without disclosing the values.

Please show me the part in the TOS where it says "any schematics posted must have values".

Again, I am only interested in showing less common alternatives to not getting the desired results. And no, I don't do it to be famous.

I'm done.

Good. I have been done for weeks, but you insist on taking the same line of thinly disguised NIHS and Ad Hominem anyway.

I will put you on ignore. I value my time too much too waste any more of it on this.

Thor

 

[Tfive]

I show what I show and I expect that those who understand electronics understand what I show and those who don't who do not, do not care enough to make themselves knowledgable.

Why would you imply that folks (not yet) understanding fully what you say do not care to make themselves knowledegable? Part of my education - still - is through this forum. Might take me much longer time and sometimes seemingly stupid questions arise to the knowledegable. But in the end I'm learning almost every day.

I get it, you don't want to be a teatcher and/or you don't have time for that. But outright insulting people with less knowledge than you have - not the way to go IMHO.

 

[Markw4]

Not everything you give has to come from something you get back. I feel very sorry for you!

Again you jump to conclusions. I am not in favor of transactional relationships. Overall it can be a very bad thing. For the purpose of my last post it may be considered as a grossly oversimplified model used for rough illustrative purposes. I can't go off on a whole different and very wide topic of relationships and organization is a simple post about the merits of a listening panel.

That said I know a guy who makes products that cost thousand of $$$ each to build. He doesn't make enough money from them to give one away to each of his friends. His situation may be different from yours. Again, please try not to jump to conclusions. Better to ask if unsure.

 

[Tfive]

Overall it can be a very bad thing. For the purpose of my last post it may be considered as a grossly oversimplified model used for rough illustrative purposes. I can't go off on a whole different and very wide topic of relationships and organization is a simple post about the merits of a listening panel.

Maybe you better had explained that in the first place. I just reacted to what you posted. But all of that distracts from what I wanted to say in post #2522. Will shut up now, maybe some fruitfull discusssion will evolve around this post.

 

[ThorstenL]

How about all the brilliant minds posting recently - put together your strength of force and try to come up with some more things we can actually measure, an addition to the FFT measurements?

If you want cat skinned and you get five expert cat skinners on the job the cat will walk unharmed, as there is no way they can agree on the way to skin the cat.

Now assume you hire multiple experts in looking for a very small black cat in a very large room and you tell them - put together your strength of force and find that cat. Understand me, the is definitely there, but is small and black and in a very large dark room. And a cat.

A dish of Tuna may help in the cat case, to bring the little black cat out of the large dark room... I have yet to find a cat that can resist juicy gourmet grade tune fillet.

Is that scientific? Does it find the cat, or rather, does it change the rules of the game to make it winnable? Kobayashi Maru?

Why are new measurement procedures kind of important?

They are not important at all.

What is important is understanding. All else derives from this. John 8:32

not all new and upcoming builders have time and ears to qualify their builds.

Buy something OTS. Many of my own commercial designs are probably way cheaper than making a DIY design of equal performance.

The only reason I see for DIY is the sake of making sh!t or to get something we want and cannot obtain commercially.

Take me, for some part, into this camp. If you, the experienced developers of electronic circuits, could team up and shed some light on how modulator noise in conjunction with non ideal analog circuitry after the DAC could be quanitified - a lot of gain for the whole communitiy could arise.

I suggested it. Noise loading. It's a VERY OLD technique to tease out exactly this sort of thing.
I never succeeded to pass such a signal through any PCM2DSD converter.

But I had limits on time and budget and as long as it passed music, does it matter it cannot pass a valid test signal (Golden Sound suggested it does, in his criticism of MQA).

We cannot at this point agree if not passing a "non-musical test signal" but passing music counts as "fail".

the end, if we - very therotically - would create a better DAC that every one can build easily

It already exists. It's called integrated circuit. I like TDA1541, PCM63/1702, AD1862, most of the hybrid BB DAC's.

They are all very old, new is not always better. Ask me one day about when I met the head of Audio ADC/DAC design at a USA based chip vendor at all places in the queue at the Public Security Bureau in Peking a few weeks before the 2008 Olympics.

Among "new" chips I rate one developed by Wolfson in the UK, now marketed by a USA Chip Vendor who seem to have entirely abandoned their own tech and IP after acquiring Wolfson.

The part is called CS43131 etc. It has a datasheet thicker than the koran and if you REALL want to get the sound quality the part is capable of, it is REALLY necessary to read and comprehend ALL of it.

isn't that the whole purpose of this forum, provide the best sound possible to anyone who wants wo achieve it?

I was under the impression that it is either about sharing what one does freely (rare now, that's how it started though, like decades ago) or making money of "group buys" (more common) and to provide free business support to cheenese kopy kats.

I get the feeling that those actually interested or concerned with sound quality make up a tiny minority.

But that is all fine. In the end it seems people get what they want (if not what they need) and those who do not elect to move on.

"Waking up is unpleasant, you know. You are nice and comfortable in bed. It is irritating to be woken up. That’s the reason the wise guru will not attempt to wake people up."

"I hope I’m going to be wise here and make no attempt whatsoever to wake you up if you are asleep."

"It is really none of my business, even though I say to you at times, "Wake up!" My business is to do my thing, to dance my dance. If you profit from it fine; if you don’t, too bad!

"As the Iraqi Marsh Arabs used to say (before being exterminated by Saddam for not being enthusiastic fans): 'The nature of rain is the same, but it makes thorns grow in the marshes and flowers in the gardens.' "

Thor

 

[ThorstenL]

Why would you imply that folks (not yet) understanding fully what you say do not care to make themselves knowledegable?

I did not imply anything. If you want to learn, you will. But I'm not available as teacher, sorry.

In my experience very few people want to learn how to fish. They are just like cat "give me the fish now" (we have two of them).

Part of my education - still - is through this forum.

I would suggest that it at best rates supplementary?

Start with Horrowitz & Hill. That is EE101. The take D. Self, Bob Cordell, JLH, Olson, Shorter and Geddes, that is Audio EE101. For digital audio (not just) there are Meitner, Dunn, Hawkesford and not the MvG.

Buy all Issues of Linear Audio AND READ THEM.

Might take me much longer time and sometimes seemingly stupid questions arise to the knowledegable. But in the end I'm learning almost every day.

You are learning 5th or more hand and doing so is highly inefficient in most cases.

I get it, you don't want to be a teatcher and/or you don't have time for that. But outright insulting people with less knowledge than you have - not the way to go IMHO.

Stating a fact used to be stating a fact.

Since when did it become an insult? Is stating that black is not white an insult to the colour black? Is it suggesting that black should be white?

I'm from the mountains. Sometimes in winter we have a mix of fog and snow that makes everything wite. It is quite lethal to get caught in it. There is no shadow, no shades of grey. Everything is white, like inside a cloud in a plane. Without the plane. Somewhere up on the mountain. If you move, you see no path and risk falling and injuring yourself (or worse), if you don't move, you freeze.

Quite a state to be in.

Thor

 

[MarcelvdG]

OTOH, a shift in the noise floor and or a shift in spectral line noise skirt width looks small on an FFT. And there is no published research on human sensitivity to such effects. Thus many people feel it is justified to assume such effects are completely negligible.

Analogue tape recorders cause a considerable random modulation in phase (wow and flutter) and amplitude (due to variations of the tape, drop-outs are an extreme case) of the signal, and therefore cause noise skirts. The effect of that must have been studied well decades ago.

 

[ThorstenL]

I release all my designs as open source/open hardware.

I used to. I mostly stopped when I found after spending a lot of time helping who I though was an individual to diagnose why their build did not work.

After a fairly painful process (not helped by language barriers) we nailed it down to a fake 2SK147 J-Fet, long obsolete by then which was not even a J-Fet.

That was not the problem. The problem was that at that point I got an email asking me to help to help fix the design to work with the part that was in there. I suggested to just buy genuine parts, to be told :
"We already have made 2,000 for a customer and need to deliver. You must help us, we invested our money into making your design. We cannot buy those parts, they cost more than our selling price!"

Seeing my public domain circuits in very expensive western High End gear passed off as "something we developed with a lot of time and effort" did not help either.

So I will only release things in the public domain that either "illustrate principles" or that use at least one widely faked and completely obsolete critical part, preferably several.

BUT, I am happy to share the principles I use, which will benefit (I hope) those who understand them,.

Thor

 

[ThorstenL]

Analogue tape recorders cause a considerable random modulation in phase (wow and flutter) and amplitude (due to variations of the tape, drop-outs are an extreme case) of the signal, and therefore cause noise skirts. The effect of that must have been studied well decades ago.

It has been studied, including audibility.

But given the systems are radically different, there is no direct applicability.

If we so want, "Jitter" in an audio sense is just "wow and flutter" in miniscule amounts. But the modulation frequencies etc. et. al are often radically different, as is audibility, even if FFT skirts look similar.

The issue is, if you carry a hammer, every problem looks like a nail.

Thor

 

[MarcelvdG]

Discussed, yes, in way. However we still end up looking at a lot more PSS FFTs summarized as THD and THD+N in this forum than any other consideration. Why? Even when some recent FFTs of the output stage were posted, in that moment, those distortion numbers felt more important than anything else. Did you feel it too?

Recently we looked mostly at distortion that is not harmonic and has a weird relation to the audio level.