[pjchappy]

Last warning:  stop the pissing contest or I am trashing this thread.


Paul

[wakibaki]

YOURSELF keep on advertising a circuit with wrong input impedance..

Keep on advertising? I once relied in error on the expertise of an individual widely respected for his work.

This makes you right in this instance and me wrong? It is this kind of sloppy thinking and partisan argument that makes me question your conclusions. If your intent is to rely on arguments such as this,

1 You have no interest in discovering the truth

2 You will convince only those with no forensic intelligence.

Here I'm using my home SPDIF cable (2.3m)- real situation -, Hiface as driver, and the scope is in place of the DAC receiver input.

As usual you show a picture of something like the arrangement in question, but not the same. There is no DAC in the circuit. Why not? Why is there so much crap on the top of the signal if it is terminated in 50R at the scope? Where is the reflected edge @ ~23nS? What is the relevance of this picture?

Can't you capture a picture that shows a single pulse?

Count the nanoseconds, this is a 192kHz signal, the primary reflection is smack in the middle, but there is still life going on at 88nsec as well!

Show me where I referred to 192k.

88nS. Show me where you warned readers using 44k1 and 48k that they would in all probability see no effect.

You have not responded to the issue concerning a non-linear rise in jitter as signal levels become marginal for DAC lock-up.

I have suggested to you elsewhere, and I repeat the suggestion here, that you should produce eye-diagrams of the recovered clock showing a credible audible improvement in jitter, or before-and-after FFTs showing a credible audible difference. Eye diagrams are not that difficult to produce, you need a scope with persistence. Your preference for producing scope shots from which much can be inferred, but nothing proved, leads me to believe that you have no confidence that you can demonstrate an audible effect.

What really bothers me about this is the information is presented entirely without caveat.

'Put an attenuator in your SPDIF cable and as long as the DAC locks up it will improve the sound' is basically what you have said.

You think you can hear a difference, but you do not have sufficient self-doubt to expose yourself to an informal blind test, or if you have, you haven't mentioned it anywhere that I have seen.

Get somebody to swap the attenuator in and out a couple of dozen times out of sight of you, identify whether it's in or out successfully, come back and tell me that you really can hear the difference, and you'll make more impact than you'll ever do with scope shots.

w

Edit:- thinking about it, an eye diagram isn't going to hack it, because it doesn't distinguish between deterministic jitter and random jitter. However, you can do FFTs on the jitter with that fancy new scope of yours can't you? However you still need to show 10-20 nS to come up to the levels detected by ear in the Benjamin and Gannon study. Or show that you can hear the levels involved yourself.

[wakibaki]

IFF you ignore the reality of high speed circuit design.  Granted, many designers do indeed neglect this aspect.  Digital links should NOT be alterable by an attenuator, but poor design prevents ideality..

Please explain how inductance in the feedback loop affects the roundtrip time of a reflection.

Perhaps you should point out to the readers that placing an attenuator at the receiver attenuates both the signal and the noise.

Please explain the physics underlying the above quote, jneutron.

w

[jneutron]

Please explain how inductance in the feedback loop affects the roundtrip time of a reflection.

Why?  I have never stated this.

What I have stated, is that if the reflections are a result of passive components, the result will always be timewise linear.

I have also stated that if the reflections are a result of an active node bounce due to amplifier slew limitations, then the output of the amp may act in a non linear fashion.  edit:  In addition, I have stated that how far the amplifier can go into saturation will be a function of the inductance of the feedback loop.  That inductance prevents the amp from zeroing the input node as fast as the input can slew, so the amp can actually saturate one or two stages internally..  Getting the stages to come out of saturation takes time. If you examine the two drive cases shown by Josephk, you can see high drive taking about 10 nSec, low drive taking 6. (and, this is only looking at the node recovery, not the output recovery after the node has settled.)  If you plot recovery time vs slew rate, you will find a specific value of slew rate that does not require recovery. Also, until the output can zero the node, the output will continue it's upward climb until the input is forced at or near zero, where it will then start to drop..  That is non linear operation, and certainly not operation intended by the chip designer.  And, it is the waveform I asked JK to see if he could capture for us to examine.
I have also stated that from the examples given by others, the recovery of that amplifier can indeed be within the timeframe that a short cable will be capable of returning a bounced signal.

Please explain the physics underlying the above quote, jneutron.
w

Why?  Perhaps you should ask the author of that quote.

Cheers, John

[wakibaki]

@ jneutron

So it's your intent only to disagree with minor points in my argument?

Otherwise respond to my post #161 ...main thrust onwards.

w

[jneutron]

@ jneutron

So it's your intent only to disagree with minor points in my argument?

I have been responding to your posts where you fabricate statements and claim I've made them.  Like your just asked question about how the fb inductance affects the rountrip time of a reflection.  Since I've never claimed such, for you to fabricate erroneous diversionary questions is a waste of time..

Otherwise respond to my post #161 ...main thrust onwards.

Respond to what?  Was there a question?  You did write a whole lot of stuff.

Cheers, John

[wakibaki]

I have been responding to your posts where you fabricate statements and claim I've made them.  Like your just asked question about how the fb inductance affects the rountrip time of a reflection.

With this in mind, consider what the output will look like if R2 is inductive.


Where will that second spike be in relation to the initial waveform?

Inquiring minds want to know...

And the fact is that it makes no difference whatsoever to the arrival time of the reflection.

IFF you ignore the reality of high speed circuit design.  Granted, many designers do indeed neglect this aspect.  Digital links should NOT be alterable by an attenuator, but poor design prevents ideality..

I guess I'm talking about the arrival time at the input, and you're talking about the arrival (!) at the output. But 'fabricate statements'... I think you are overegging the pudding.

w

[jneutron]

I guess I'm talking about the arrival time at the input, and you're talking about the arrival (!) at the output.

As I stated before, you are looking only at the reflections at the input, but others have been considering the overall effect on the system.<===  I changed the font to emphasize the specific point I have been repeating over and over..

As I stated before, the input attenuator, when applied to an active node input, will reduce the non-linearities of the stage caused by a slew rate beyond the circuit's capability.

As I stated before, any reflections caused by an active node which make the double transit and arrive back at the active node, the leading edge of that bounce will propagate directly through the amp.

The net result will be an uncertainty in the leading edge.  My understanding of spdif is that the clock is recovered using a pll, and that jitter in that clock will arrive at the audio out as noise.  I cannot however, answer the question:  is this jitter caused by active node reflections caused by excessive input slew rate sufficient to become audible..

The bounce, and the resulting uncertainty in the leading edge, will go away if the attenuator causes the input signal at the node to lower it's slew rate sufficiently to stop non linearities.

But 'fabricate statements'... I think you are overegging the pudding.
w

One only has to review all your posts to me to find multiple examples of such.

It's like that doctor/patient exchange...

Doc, it hurts when I do this..

Well then, don't do that!

If you expect civil dialogue, act accordingly.  Your tech stuff is nice, but ya mess up the presentation big time..

Cheers, John

[wakibaki]

John,

You have made numerous personal comments about me. For the large part my posts have been technical in nature.

w

[JohnR]

wakibaki - you may remember that I did not approve your very first post. Every post of yours is laced with challenges and putdowns. Kindly up your game. Thanks.

[jneutron]

John,

You have made numerous personal comments about me. For the large part my posts have been technical in nature.

w

I have stated that your presentation is terrible.  Everybody here seems to think the same way.

I wish for you to slow down and read what I've posted more slowly.  Only now have you understood I'm talking about the system, not the input line.  That was the square peg in a round hole comment.

Stating you need to be civil is consistent with your own statement that you need to work on your people skills..

Cheers, John

[wakibaki]

Oh. OK.

w

[rollo]

 how would a transformer [ pulse trannnie] coupled digital input help this issue ? Or is that for another pupose ?


charles
 

[art]

HA! This thread still exists!

I had completely forgotten about it. And most of us are probably thankful for that.

Anyway, since I have a few moments on my hands, before I head back to the salt mine, let my give a history lesson, of sorts, on how this attenuator stuff came about.

Seriously, it is one of the oldest, and possibly worst-kept secrets, in the RF world. I do not know how or when it got started, but I can talk about microwave radios, that were used in the telecom world, for many years. (You know, back in the good ol' days, when Ma Bell gouged everyone on long-distance calls, and you had to wait until the weekend to call Uncle Bill, to thank him for the birthday present he sent you. And were only allowed a certain amount of time, because it cost so dang much. You could call it the "analog" era, because there really wasn't any form of "digital" communication.) (Ok, technically, Ma Bell had "T-carrier", but that was only for local use. And it had a very distinctive quantization noise. Long-distance used "L-carrier, which in Ma Bell's case was 4 GHz microwave radios. Some of you old timers may remeber those strange red and white HUGE towers, just off the Interstate highways, with the odd-looking "cornucopia" antennas.)

So, back to  the story................

It was very common to insert a small pad, maybe only 2 or 3 dB, after the mixer, and in front of the IF preamp. You didn't have to worry about any sort of post-mixer image filter, because the image would have been "RF GHz + 70 MHz", as the gear used 70 MHz IF stages. But, you did need to make sure the input of the IF preamp had a decent "return loss". (Fancy way of saying the impedance had to be a close as possible the right value, and wasn't mainly a reactive component.)

The problem was the IF preamp usually had some sort of delay equalizer built into it. Which meant that it contributed to the input reflection, and in a nasty way. The delay equalizer also had something called "group delay", which was unavoidable. The problem was when the group delay looked nasty, it gave rise to something called AM-PM conversion. (Some of you hardcore amp designer types might recall a lot of hand-wringing and food fights, about how that can muck up the sound of an amp. But, let's steer clear of that!)

Since AM-PM basically caused IMD to increase, that would affect how many telephone channels you could load onto the system before it sounded so bad it was unusable. The problem was made worse when the impedance presented to the mixer looked nasty. Which was largely unavoidable, since you really did need to have that delay equalizer in the chain. (We'll skip why that was needed, because that is not the point. The point was it was there, it wasn't going away, and someone had to find a way to mitigate it.)

So, this is where one of the worst kept secrets came into play. Darn near everyone (at least in the RF world) knew attenuators could "fix" reflection problems. The obvious drawback was decreased S/N ratio. Which translated to a loss of "fade margin". IOW, how low the signal could fade to where the system became unusable.

So.................for EVERYONE who built microwave telecom radios, to employ this trick, to limit AM-PM conversion, at the risk of lower SNR and fade margin, it had to make s significant difference.

It did. Which is why everyone did it.

"Yeah, yeah, yeah, makes for a great story, grandpa. How in the world does that related to digital audio? I mean, there is a spec and all, so why do you think you can ignore it?"

No one is ignoring it. But, you would be surprised on how many folks who build this stuff are really oblivious to how it works.

(I know this will sound condescending to some, but I can tell you, after 40+ years of doing this "high-end" stuff, that most "audio designers" have no clue how "RF" works, and treat RF the same as AF, and really get a lot of it horribly wrong. Trust me, even if you hate me.)

Let's stick to SPDIF, as the AES/EBU standard is really to vague to merit discussion. OK?

The spec calls for 0.5 V, p-p, into a 75 ohm load. (Also implies the output impedance is also 75R. HA!!!!!!!!!!!!!!!)

Wanna guess how close ANY transport or DAC is, to 75 ohms? Well, I can tell you.............

There are TONS of products with a return loss in the 10 dB range. Which means the reflected signal is around 30% of the transmitted signal!  Which also translates to an impedance of either (roughly) 39 ohms or 144 ohms. No so good, right?

So, there is problem #1.

Next is the even easier problem: the wrong level.

Yes, hard to believe there are folks in this business that have no clue how to even get the voltage right. They just stick a 75R resistor in series with whatever drive chip they have, and off we go! Not only is the level going to be wrong (for obvious reasons), but even the impedance is wrong!

The "audio engineer" types forget to take into account the intrinsic impedance of the driver chip. That is a function of the process the chip is made with. (Should not be a surprise, there, right?) The problem is the old CMOS stuff was around 28 ohms or so, and the fancy high-speed stuff of today is only a few ohms, but it is still there. And should be factored in. But isn't..............

(Ok, another famous snafu, from days long gone.............a certain "mod company" used the old "insert 75 ohm resistor here" trick, in one of their products. Since they were big in the aftermarket gimmick bidnis, they had to come up with a magic cable, that you simply had to buy! And their favorite pet reviewer, on cue, wrote a glowing review on it, and pimped it to death. Just one problem: these guys could obviously hear, as they came up with the right cable.

For their flawed system. Yes, they discovered a 93(!) ohm coax able that sounded right in their system. Not so when you played "mix and match" with other components. The dealers were confused.

"And I bet you didn't lift a finger to help them. You guys probably sat back and laughed your butts off."

How well you know us!)

So, take a transport, with a drive impedance that isn't quite right, and may or may not be too hot, drive-wise, and a DAC with......................uh, let's forget about how bad DAC input stages are, because my fingers (and probably your eyes) are getting tired. And what do you have?

A mess! Yes, it works. And works well enough that folks buy this stuff and are happy. All fine and well.

But, if....................for the cost of a few resistors (we'll neglect the cost of the mechanical assembly needed to house this and actually make it work, unless you take your gear apart and build into it), then what is the harm in trying?

OK, some systems may actually stop working. Given what I know, about the chips used inside (well, at least the ones in the past, and not the newer stuff which uses one chip to do the receiver function and the "state-machine" decoding, to get around those nasty PPLs us RF guys love to jack with) you would be surprised how much you can attenuate the signal with no problems. Let's say I have yet to see one that could not take 10 dB shoved in front of it, somewhere in the chain. Either all at either end or broken up between the 2 ends.

Lower S/N? Yes. Less jitter created at the interface? Yes. Which is more important?

You be the engineer.

I've given you the rationale, and you have to decide which way to turn. (Probably the way that sounds best, right?)

[audioengr]

Wanna guess how close ANY transport or DAC is, to 75 ohms? Well, I can tell you.............

There are TONS of products with a return loss in the 10 dB range. Which means the reflected signal is around 30% of the transmitted signal!  Which also translates to an impedance of either (roughly) 39 ohms or 144 ohms. No so good, right?

99% of S/PDIF outputs are not matched well to 75 ohms.  99% of all S/PDIF inputs are well-matched to 75 ohms.  All this takes is a resistor usually.

Return loss is of little importance.  This is not radio.  It is the reflection and the reflection timing that is important.  It's the signal integrity at the receiver that's important.  You know that..

Lower S/N? Yes. Less jitter created at the interface? Yes. Which is more important?

Excellent question  I wish there were DB tests showing exactly this.  This is the problem I have with the guys at Audio Science Review forums.  They say the measurements are the only truth and you cannot trust your ears, but at the same time, they cannot say whether it's more important to have low jitter, low HD, low noise floor or low crosstalk.  The audiophile community needs this work to be done.  Give each of these measurements a weighting based on actual human hearing.

Only then can we maybe look at measurements and say with any certainty that one component will sound better than another.

BTW, do you believe that faster edges are the ticket to lower jitter? I won't tell you what I think yet.

Steve N.

[art]

99% of S/PDIF outputs are not matched well to 75 ohms.  99% of all S/PDIF inputs are well-matched to 75 ohms.  All this takes is a resistor usually.

Maybe you haven't measured as many as we have. No, they really are bad.

Easy in theory; poor in execution.

Return loss is of little importance.  This is not radio.

Another disagreement. We have lots of measurements, with correlating listening tests that showed what the RX return loss needs to be in order for the listener perceive all cables as sounding the same as any other one.

(Key word is "perception". Something that is lost on some folks.)

And since all cables don't sound alike, it allows some of us to make a living selling them. (We do not advertise ours, as we view it mainly as a service to our customers. Manufacturing and marketing are not our forte. Obviously.)

It is the reflection and the reflection timing that is important.  It's the signal integrity at the receiver that's important.  You know that..

And return loss doesn't play a part in that, eh?

Excellent question  I wish there were DB tests showing exactly this.  This is the problem I have with the guys at Audio Science Review forums.  They say the measurements are the only truth and you cannot trust your ears, but at the same time, they cannot say whether it's more important to have low jitter, low HD, low noise floor or low crosstalk.  The audiophile community needs this work to be done.  Give each of these measurements a weighting based on actual human hearing.

Only then can we maybe look at measurements and say with any certainty that one component will sound better than another.

Not familiar with those folks, but I suspect we could not get along with them. In our limited experience, folks like that have a preconceived notion of what works, what doesn't, and why. Are not open to hearing dissenting views.

The one thing that most of the meter readers can not do is relate how doing x to a musical passage will  affect it, in what manner, and how it will be perceived by the listener.

Actually, there are 2 areas where we have tons of "data" on. Already mentioned one. The other is clock jitter. Which is much more controversial, because no one wants to hear what goes on below 1 Hz is important.

BTW, do you believe that faster edges are the ticket to lower jitter? I won't tell you what I think yet.

Steve N.

That is a ambiguous question. Jitter means different things to different people. When we talk of "jitter", we are strictly talking about clock jitter. And I am not going to discuss that subject. My patience is long gone, on that matter.

When you have other design consultants and/or manufacturers say that no one can believe your results because................well, probably because we are in Texas, and all that you need to hear "clock jitter" is to modify an FM radio to listen to the signal, it is pointless to continue.  That is where I am at.

The problem is everyone wants their clock to sound like one that costs $100, but still only costs $1. Just because we are stupid enough to buy 1000 pieces of mystery parts, and take the time and spend the money to pick through them, to find the "diamonds in the rough"....................well, you would be surprised how much we end up paying for them.

At least at the end of the day, if we have a $30 part, it really does measure like the ones that actually do sell for $30 are supposed, only a helluva lot better. (I say "supposed", because....................well, guess what else we have data on.)

[audioengr]

That is a ambiguous question. Jitter means different things to different people. When we talk of "jitter", we are strictly talking about clock jitter. And I am not going to discuss that subject. My patience is long gone, on that matter.

The jitter that usually matters, jitter on S/PDIF or on I2S to the D/A. This is what I measure here:

https://www.audiocircle.com/index.php?topic=157348.0

and here are the measurement fanatics using a DAC for the same measurement (I prefer direct measurement for sources):

https://www.audiocircle.com/index.php?topic=163027.0

The problem is everyone wants their clock to sound like one that costs $100, but still only costs $1. Just because we are stupid enough to buy 1000 pieces of mystery parts, and take the time and spend the money to pick through them, to find the "diamonds in the rough"....................well, you would be surprised how much we end up paying for them.

I get this.  The thing I focus on is not so much the oscillator, but all of the associated dividers, muxes, registers, selectors and buffers. These can take a 1psec phase-noise oscillator and make the S/PDIF output or I2S output have 800psec of jitter.  I have developed a bag of jitter tricks over the last 23 years that is very effective.  I still get custom low-jitter oscillators that may cost $25, but I don't bother with oven-control or rubidium clocks because all of the other logic will usually add significantly to the jitter. I figure a factor of 5-10x the oscillator spec on the bench is what you actually end-up with on I2S, best-case and the power noise and ground noise factors in as well.

[art]

The jitter that usually matters, jitter on S/PDIF or on I2S to the D/A. This is what I measure here:

https://www.audiocircle.com/index.php?topic=157348.0

and here are the measurement fanatics using a DAC for the same measurement (I prefer direct measurement for sources):

https://www.audiocircle.com/index.php?topic=163027.0

Oh, ok...............that kind of jitter. Got it.

I get this.  The thing I focus on is not so much the oscillator, but all of the associated dividers, muxes, registers, selectors and buffers. These can take a 1psec phase-noise oscillator and make the S/PDIF output or I2S output have 800psec of jitter.

Here again, we are talking apples and oranges. The "jitter" of those clocks, the published specs of those clocks, are really meaningless, when it comes to digital audio.

Those clock specs are really nothing more than a measurement of the noise floor. If you look closely, they always say "jitter frequency >1 kHz". IOW, the noise floor. If you measure the "jitter" using one of those fancy megabuck 'scopes, they use the telecom mask, which if I am correct is from 12 kHz to 100 kHz. Again, more nothing. Noise floor is meaningless.

I have developed a bag of jitter tricks over the last 23 years that is very effective.  I still get custom low-jitter oscillators that may cost $25, but I don't bother with oven-control or rubidium clocks because all of the other logic will usually add significantly to the jitter. I figure a factor of 5-10x the oscillator spec on the bench is what you actually end-up with on I2S, best-case and the power noise and ground noise factors in as well.

Oven control is not needed. Even for a $100 SC-cut crystal. (Yes, it drifts like mad. Sounds incredible. Again, I have been told that is impossible, but we have done it.)

Don't get me started on rubidium clocks. TOTALLY useless for audio.

As for your "tricks".............if they work for you, then I am fine with that. I'll just add if you think you are really getting your money's worth, for your $25 "low jitter" clocks, well, how 'bout this..................

A one time and one time only offer.

Send me a few. I will measure them. Gratis. We usually charge close to 3 figures to do this, for other companies. The only caveat is that I can post the results here, and make sure the whole world sees them.

BTW, all the crap added from muxes and dividers and other digital sources of crap have ZERO effect on what happens in the critical jitter range. Mind you, I am not saying it is not important or that you can not measure it or that it won't affect the sound. What I am saying is that the really important stuff (you know, down below 10 Hz) is unaffected by it.

This is precisely why crappy ol' SPDIF has any chance of sounding good. Even with a noise floor that is 30 dB higher than the source, eventually the phase noise of the recovered clock tracks the source clock. Which is why we can demo our gear with a ratty old DAC that was made almost 30 years ago, and you can hear a good clock from a really good clock.

I know.................shocking, isn't it?

[audioengr]

"BTW, all the crap added from muxes and dividers and other digital sources of crap have ZERO effect on what happens in the critical jitter range. Mind you, I am not saying it is not important or that you can not measure it or that it won't affect the sound. What I am saying is that the really important stuff (you know, down below 10 Hz) is unaffected by it."

I'm not sure what the critical jitter range is yet.  I have yet to see ANY correlation of jitter spectrum measurements to listening tests  The closest I have come to correlation is the jitter histogram that shows the distribution as well as the spectrum of the same.  I have seen some correlation there, but it's inconsistent.  When you have distribution, magnitude and spectrum all factoring in, it is hard to derive solid conclusions.  A LOT more study needs to be done in this area.

This is precisely why crappy ol' SPDIF has any chance of sounding good. Even with a noise floor that is 30 dB higher than the source, eventually the phase noise of the recovered clock tracks the source clock. Which is why we can demo our gear with a ratty old DAC that was made almost 30 years ago, and you can hear a good clock from a really good clock.

My own DAC, by design, has no reclocker on the S/PDIF input.  Direct to the receiver.  The best DACs avoid reclocking on the inputs IMO, like the Metrum Acoustics DACs.  This way they get the benefit of truly low-jitter sources.

Using the optimum S/PDIF receiver is critical though.  They are not all the same.  Some of them can result in much lower jitter to the D/A, others are junk.

[audioengr]

A one time and one time only offer.

Send me a few. I will measure them. Gratis. We usually charge close to 3 figures to do this, for other companies. The only caveat is that I can post the results here, and make sure the whole world sees them.

They are 5X7mm surface-mount.  Can you deal with that?  I may be able to send you a SILABS 530CB49M1520DGR  - this is 49.1520MHz 3.3VDC  It may have solder on the pads.  Period jitter RMS spec 2psec. After my logic is added, 7-10 psec is the result.

Steve N.