I really don't see why it has to be shown that the reflections fall on the transition - surely this is just a matter of the return timing of reflections? The fact that there are reflections means that they can give rise to jitter under the circumstances that cause them to interfere with the rising or falling edge. So anything that can be done to reduce these reflections must be of benefit, unless of course you are sure that your reflections are not falling on your transition edges? Am I missing something?
Again, I put out the plea for others to test the premise in the real world & report back!
Again, I put out the plea for others to test the premise in the real world & report back!
Last edited:
Can I ask Joseph K - the "direct no SPDIF" (properly terminated) scope shot seems even better than the "cable + attenuator" scope shot - any room for improvement here using an attenuator on "direct no cable" connection?
The overshoot in both scope shots is not the small wiggle you are talking about I presume? Is this from the Hiface transformer or the driver, do you reckon? Is the samll wiggle of significance & any idea what is the origin of this?
I am unable to see that you have demonstrated anything with these screenshots other than a muddle of overlaid images.
You have asserted that you are showing a reflection coincident with a falling edge, but the slopes of the 'reflection' are so shallow, and the amplitude of the 'reflection' so small, that it's impact on the the timing can only be inconsequential.
What I want to see is the SUPERPOSITION (your emphasis) of the reflection causing appreciable movement in the distance between the rising and falling edges, before and after attenuation, as measured with a cursor. I also need to see that this variation is greater than the intrinsic jitter generated by the transmitter.
You will not be able to demonstrate this. If you could do so, you would have done so already. Your whole case is built on a preconception which is distorting your interpretation of the evidence.
I still fail to understand why you have not shown the upper AND lower limits of the trace.
The simple fact is, that even if you could demonstrate such an effect, you would be obliged to shorten the cable progressively until you obtain the coincidence that you claim is commonplace. The likelihood of anyone's cable randomly being the exact length required makes a nonsense of the idea that a significant number of people will see an improvement as a result of following your suggestions.
w
You have asserted that you are showing a reflection coincident with a falling edge, but the slopes of the 'reflection' are so shallow, and the amplitude of the 'reflection' so small, that it's impact on the the timing can only be inconsequential.
What I want to see is the SUPERPOSITION (your emphasis) of the reflection causing appreciable movement in the distance between the rising and falling edges, before and after attenuation, as measured with a cursor. I also need to see that this variation is greater than the intrinsic jitter generated by the transmitter.
You will not be able to demonstrate this. If you could do so, you would have done so already. Your whole case is built on a preconception which is distorting your interpretation of the evidence.
I still fail to understand why you have not shown the upper AND lower limits of the trace.
The simple fact is, that even if you could demonstrate such an effect, you would be obliged to shorten the cable progressively until you obtain the coincidence that you claim is commonplace. The likelihood of anyone's cable randomly being the exact length required makes a nonsense of the idea that a significant number of people will see an improvement as a result of following your suggestions.
w
This would take months of testing setup and catch it causing a problem, there is a certain amount of inductive reasoning implied.
Are you arguing that jitter is never caused by reflections? Can you prove that?
Firstly, you now seem to accept that there are reflections & that the attenuators reduce these reflections, or am I wrong in this? This is a significant shift in your position from the start of the thread. So have these plots shown you this or how exactly did you come to this realisation?
Again, I'm not being sarcastic, I just need to understand your logic.
Now you want to be shown that these reflections can fall on the rising edges but surely you don't need this demonstrated as it can be deduced! You seem to be saying that this is a very rare occurrence & as a result a trivial case - can you prove this?
Last edited:
No, I am not arguing that jitter is never caused by reflections. In fact it is. I can point you to a paper on the subject, however the medium is 50 ohm microstrip on an exotic substrate (fr4 is not much good over 2.5GHz) , the clock frequency is 10GHz and the bitrate is 5Gbps. And anyway, it's not my intent to make that case, but it sure shows how ill-researched the protagonists on the other side of the argument are.
We're talking here about common old co-ax, lumped components and a bitrate that barely gets into the Megas, if at all.
Which is how I've known from the beginning that the work simply hasn't been done, and that the whole thing is just based on 'inductive' reasoning, or, more accurately, 'blind faith'.
What I am saying is that these plots do not demonstrate that it is occurring in this case, that the suggestion that significant mistermination is commonplace (on which this whole thread is predicated) is is totally without foundation and that the purchase of RF attenuators for the reduction of jitter is, in this case, a waste of money, time and effort. Even if you think you can hear a difference, that's not the same thing as a blind test conducted with a statistically significant number of people.
85 posts later and no-one has produced any evidence worth the name. No evidence. Let's face it, you haven't got any.
w
Wow! So you are saying that reflections only cause jitter above a certain MHz range? Can you say what that range is or where it starts?
So even if a statistically significant number of people report that there is an improvement you will reject the jitter argument - OK!
No evidence according to you! I believe reflections are shown. I believe that the attenuator reduces these reflections are shown. Let the readers make up their own minds on this.
Why? Why wouldn't a blind test with one person and a significant number of trials be valid?
If you're talking about anecdotes rather than the results of controlled listening tests, then indeed there's no reason to accept that, any moreso than to accept the existence of ghosts, the Loch Ness Monster, or alien abductions. The concept of "statistical significance" is predicated on actual data.
I find it especially amusing that wakibaki does not like the shots taken of a setup assembled using a real SPDIF driver & cable and connectors. The only difference to my home system is the scope in place of the dac input.
Maybe he has some problem with practical experiments? Real life signals which are always complex & dirty & difficult to decipher?
But staying on topic: my original point was that an inserted attenuator helps with the reflections in the line. This all originates from Jocko Homo.
May we consider it now approved?
The next (continously evolving, in defense of what?) point is if these reflections are causing additional jitter or not. I had already told that the effect is subtle. I had shown that the coincidence of these reflections and the transmitted edges can be a real life situation.
I had tried to explain, why superposition of the original & reflected signals can cause a distortion of the transmitted edges. We are talking about the order of ten -hundred picoseconds, as already told before.
Detecting this in a real life situation is not easy at all, and wakibaki is trying to play this card. I think I had done my contribution up to now, maybe now he could show us something, apart from academic superiority?
Would like to emphasize, that showing proof of this type of jitter was not the original claim at all. Not because it does not exist, only because it's not easy at all.
By the way,
It took me ~half an hour to assemble a demonstration of the effect of reflections on the falling edge of the previously used HP generator. The scope is a good old 7104.
The first shot shows the generator signal.
the second shot is where I had deliberately introduced strong, fast reflections so that they partially fall on the original edge. Note that "strong" is not a big amount anyway!
Also note that the fall time of the original edge is ~1ns, close to the real life value ~2ns of the Hiface driver.
Maybe he has some problem with practical experiments? Real life signals which are always complex & dirty & difficult to decipher?
But staying on topic: my original point was that an inserted attenuator helps with the reflections in the line. This all originates from Jocko Homo.
May we consider it now approved?
The next (continously evolving, in defense of what?) point is if these reflections are causing additional jitter or not. I had already told that the effect is subtle. I had shown that the coincidence of these reflections and the transmitted edges can be a real life situation.
I had tried to explain, why superposition of the original & reflected signals can cause a distortion of the transmitted edges. We are talking about the order of ten -hundred picoseconds, as already told before.
Detecting this in a real life situation is not easy at all, and wakibaki is trying to play this card. I think I had done my contribution up to now, maybe now he could show us something, apart from academic superiority?
Would like to emphasize, that showing proof of this type of jitter was not the original claim at all. Not because it does not exist, only because it's not easy at all.
By the way,
It took me ~half an hour to assemble a demonstration of the effect of reflections on the falling edge of the previously used HP generator. The scope is a good old 7104.
The first shot shows the generator signal.
the second shot is where I had deliberately introduced strong, fast reflections so that they partially fall on the original edge. Note that "strong" is not a big amount anyway!
Also note that the fall time of the original edge is ~1ns, close to the real life value ~2ns of the Hiface driver.
Attachments
Last edited:
I'm not a protagonist by any means, anytime I save money I'm all for it. Are you saying that properly terminated at the transmitting side and receiving side reflections are a non issue or are you saying this doesn't matter either? (again not antagonising trying to understand your argument.) And yes mistakes are often made with inductive reasoning.
Where did I say that?
By all means take issue with the things that I have said.
What I will say is that reflections are more readily controlled in co-ax with lumped componemts than they are at microwave frequencies in microstrip.
w
You obviously can't conduct a mature conversion if you don't understand the meaning of a question mark.
unsubscribed.
Lumped components? I have shown in many different shots the Hiface driver's signal shape.
In a best case I have measured 1.8ns, symmetrical rise/fall edges. How much is the equivalent bandwith?
For me it's 200Mhz.
Then, why did You brought up the bitrate? What importance does it have? We are talking about EDGES here! And anyway, common old coax is what is used even in satellite receivers...
I'm not trying to help or hurt anyone; to be honest, most of what's being argued here makes no sense to me, since there's no evidence that any of this stuff is audible. Nor should it be, since the DAC process has a separate clock. But your argument in this regard was invalid. Should incorrect arguments be allowed to stand?
No, what I am saying is that properly terminated at the receiving end reflections are a non-issue. The transmitting end is irrelevant, as reflections at the transmitting end are returned to the transmitter. Once the signal enters the cable, it is completely absorbed at the receiving end, if the termination is correct. Only if the coefficient of reflection is non-zero is a reflection returned to the transmitting end, and only if the coefficient of reflection at that end is non-zero is a reflection returned to the receiver.
Suppose 10% of the signal is reflected at both the transmitting end AND the receiving end.
90% of the signal enters the cable. 10% returns to the generating device (transistor) via the pulse transformer and is normally dissipated as heat. 81% of the signal passes to the receiver and is dissipated in the resistive component of the load. 9% of the signal travels back down the cable to the transmitting end. This is UNSEEN by the receiver. At the transmitting end 8.1% of the signal passes through the pulse transfoirmer and is returned to the transmitting device where again it is dissipated as heat. 0.9% of the signal is returned to the receiver.
If (and only if) the returned signal at the receiver coincides with the rising or falling edge of the original signal it MAY interfere with the timing of that edge.
In order for the coefficient of reflection to be 10% the mismatch required is 15 ohms in a 75 ohm environment.
RG11 is specified to have an impedance of 75 ohms +-3 ohms. This corresponds to a reflection coefficient of 0.02 when terminated in 75 ohms. This is a poor quality cable, many are specified +-2 ohms.
When properly terminated at the receiving end, using a cable of average quality, reflections are a non-issue in SPDIF systems, and the inclusion of an attenuator is an excursion into the unknown, as likely to cause problems as to solve them, for users with the test equipment and level of technical expertise shown thus far in this thread.
w
Last edited:
One would have hoped so, SY, but in fact this is not always the case. Frequently DACs run off a clock recovered from the SPDIF.
If the DAC clock should be marginally different from the incoming SPDIF clock, problems can arise. If the DAC clock is slower than the SPDIF, then data piles up in the receiver. If the DAC clock is faster than the SPDIF then the data may be consumed before more has arrived.
Obviously, this is a situation which can be accommodated by buffering. But how much buffering? How much data should the DAC pile up before starting to play it? Or contrariwise, how much storage should the DAC provide to accommodate the data that is coming in faster than it is being used?
It's a lot simpler just to use the clock at the sending end, if you can make a local oscillator sync to it. This is the function of the PLL, and why the jitter rejection of this component is important.
It's a lot simpler again to put your music in a flash memory co-located with the DAC and run the whole system off a good clock...
w
Oh, and in order to detect an improvenent in quality, blind tests performed by a number of individuals are required, as distinct from the case of 'burned in' cables where the test is only for a difference, and evidence that a single individual can detect that difference is all that is required.
w
w
Simpler from the point of view of the engineering required to get a good sound, I agree. But economically this isn't really an option yet as the cost of one CD-R is around $0.1.
Compare that with TF cards which seem to be the cheapest way to buy flash storage here. $45 gets me 32G which on average stores 100CDs after lossless compression giving around $0.5 per CD. So flash memory has further to fall to make what you're proposing economic, but its come a long way since I bought my first (256Mb) mp3 player.
- Status
- Not open for further replies.