it depends , in the real world we don't deal with DC, there is always some amount of sub 1hz frequency that passes thru caps and presents itself as DC (for all practical purpose.) And all caps have leakage.
I guess what I am saying is in text books things are cut and dry, but after you close the books and start working you will learn its not all black and white in the real world. If you keep a closed mind you will never grow as an engineer.
Just because an SPDIF is terminated at 75 ohms on the receiving end doesn't make it perfect. I think you need to spend some time on the bench with a scope.
I modified a Hiface unit & the only change was to provide a clean clock supply - huge difference in sound - it agreed with all others who report the same - it agrees with the various descriptions of the sound of jitter that I've read - you even seem to know this yourself in your soundstage statement below - what else do you want to attribute this one change to?
Oh great, so you agree that it's reducing jitter & NOW you have another effect that you are attributing to it using guesswork. I've already pointed out that there are measurements of the reduction in reflections - why do you have to GUESS that something else is at work?
Oh yeah baby, it's bigger & industrial looking, so it must be better - I'll have two of them & hopefully they are very expensive so then I'll know that they are better than the $12 ones & can tell everybody so
Wakibaki,
I'm not suggesting that anybody wilfully did anything - just mistakenly made the same assumption you do & said "there's a 75R in there so our termination is perfect" - as Regal says a bit of measurement may reveal otherwise. Have you ignored the TDR plots I've been mentioning throughout all this which shows reflections in a "properly terminated" system & how these were measurably reduced when an attenuator was used? This termination was done according to the application note in the TI DIT4192 datasheet (& don't tell me that I now think I know better than Texas Instruments - just look at the plots)
I'm not suggesting that anybody wilfully did anything - just mistakenly made the same assumption you do & said "there's a 75R in there so our termination is perfect" - as Regal says a bit of measurement may reveal otherwise. Have you ignored the TDR plots I've been mentioning throughout all this which shows reflections in a "properly terminated" system & how these were measurably reduced when an attenuator was used? This termination was done according to the application note in the TI DIT4192 datasheet (& don't tell me that I now think I know better than Texas Instruments - just look at the plots)
Nothing to do with 'what I want' here - why make it into something personal? Cleaning a clock supply will also have the effect of reducing the RF content on the supply don't you think? Tell us more about how you cleaned the supply and what the cleaned supply was connected to on the board - I guess a clock osc but anything else besides that?
Its because of what I said earlier - perhaps you did not read carefully, So I'll say again - because in my experience sibilance is not caused by jitter, its caused by RF intermodulation. Reductions in reflections result in lower jitter you think? Any measurements to demonstrate that?
That's just a turn of phrase nothing else!
It went from a regulator to a battery supply but I'm not going to argue with you on this - you have your opinion that's fine!
First, I'm sure you understand that reflections arriving at the receiver during the decision window will superimpose a voltage onto the transition signal which will cause the decision point to shift to where it should be. Is this jitter? As to measurements - look at the TDR plots on DiyHiFi.org - it's plain as day!
So you cite avoidance of a non-existent argument as an excuse to avoid answering my question? Put your analyst on danger money baby
Ah but not 'plain as day' to me as I haven't been invited to join diyhifi even though I've applied. Without membership the attachments are invisible - one practical outworking of Jocko's paranoia.
So you don't disagree about how reflections cause jitter, then! If that's the case then reducing them should be a good thing, no?
I'm not sure how you got that from what I said. I'm open to being convinced but as I've seen no evidence I'll remain sceptical.
Reducing reflections in isolation I'd say would be good. However, padding the line does not do that, as Jan has already pointed out. So I remain somewhat sceptical.
Is there anything wrong with my statement of how reflections can cause jitter? If not then you agree with the theory! You just remain to be convinced of it in practise - I can understand that, I would be the same
As I already said in answer to Jan - you can't reduce reflections in isolation in this way but by attenuating ALL the signals on the line the reflections are also attenuated. Now if the signal on the line is a lot higher then spec to start with (I.e >0.5vPP) then you can bring this down to spec & your reflections will be greatly reduced in strength but your SPDIF signal will be normal. It's not that difficult to comprehend, really!
What I understood is that what is important is the ratio of the source signal to the reflected signal. If you attenuate both, would that change the ratio? And if yes, how?
I know there are 'directional couplers' which transmit a forward signal and attenuate reflections. Is this a confusion between attenuators and directional couplers?
jd
Well, if the strength of the reflected signal is brought down to insignificant levels then it doesn't matter that the ratio to original signal is maintained. I see what you're getting at & yes it is the same as correcting the termination to exactly 75ohm but I guess it's the lazy way of doing this (remember it's not just a case of putting in a 75R)!
No I'm not confusing them with these directional couplers but have you ever tried one of them in this area?
Last edited:
No, I haven't found anything wrong in principle with your statement that reflections can cause jitter. I remain to be convinced that the result in this particular incidence is indeed improved jitter. Which explains my initial question - you seemed to be basing your claim that jitter had indeed been reduced on what you heard. Which left a loophole, as potentially what you heard can be explained by other means.
Yep, so far, so good.
Then it must follow that that's evidence of incompetence in design. Why would anyone want to call something 'SPDIF' and not follow the spec?
I'm not having any difficulty comprehending this part. My difficulty in comprehension stems from the assumptions which you are bringing in beyond this. Like the one that goes 'given a larger-than-spec signal, the SPDIF receiver will perform no better than with a normal-spec signal'. Where have you imported that assumption from pray?
I had an email exchange with Jocko, and he states the following:
"Take a transmission line, and stick a short on the far end. You have a RL of 0 dB, as all the energy is reflected back to the source.
Now, stick a 10 dB pad in line, right before the short. 10 dB of the signal is absorbed, by the pad, and hits the short. Now, that signal, already attenuated by 10 dB has to pass through the pad..............again! Where it is attenuated by an additional 10 dB. So, instead of all the energy bouncing back, you only have 10% bouncing back. Which, is -20 dB. So, by sticking a 10 dB pad before the short, we have lowered the reflection coefficient, by twice the amount of the pad.
Now, if you have a rho of 0.1, instead of 1, what happens? Well, the incident wave has been reduced by 10 dB. With a rho of 0.1, only 10% of what hits the end is capable of bouncing back. Or -20 dB, of something that has already been knocked down by 10 db. Or, 30 dB.
The reflection goes back through the pad............again. (I'm sounding like Forrest Gump.) It loses 10 dB, as it should. So, now we are down to -40 dB, total reflection.
Long story short, you get twice the amount of the pad, in refection reduction."
[snip]
"Of course, we are talking about a pad with decent construction. If you make one by hand, with leaded resistors, and space-wired, with long leads, you will see a big reflection spike at its input. That will still show up on a TDR, as spikes, before the end. On a network analyzer, you would still see the 20 dB reductions, but as frequency goes up, the amount of the reflection also goes up. Which is why you would want to use a really good pad, that has been characterized for the frequency range you are working in. For SPDIF, I would use one spec'ed to 100 MHz. A hand-built one won't cut it, but would still be of great benefit."
I'm not 100% sure but it does seem to make sense. So maybe there is something in using a pad, if it is of good enough quality: having less reflection than the non-padded line to begin with. That seems to be an important condition. if that isn't the case it makes matters worse.
jd
"Take a transmission line, and stick a short on the far end. You have a RL of 0 dB, as all the energy is reflected back to the source.
Now, stick a 10 dB pad in line, right before the short. 10 dB of the signal is absorbed, by the pad, and hits the short. Now, that signal, already attenuated by 10 dB has to pass through the pad..............again! Where it is attenuated by an additional 10 dB. So, instead of all the energy bouncing back, you only have 10% bouncing back. Which, is -20 dB. So, by sticking a 10 dB pad before the short, we have lowered the reflection coefficient, by twice the amount of the pad.
Now, if you have a rho of 0.1, instead of 1, what happens? Well, the incident wave has been reduced by 10 dB. With a rho of 0.1, only 10% of what hits the end is capable of bouncing back. Or -20 dB, of something that has already been knocked down by 10 db. Or, 30 dB.
The reflection goes back through the pad............again. (I'm sounding like Forrest Gump.) It loses 10 dB, as it should. So, now we are down to -40 dB, total reflection.
Long story short, you get twice the amount of the pad, in refection reduction."
[snip]
"Of course, we are talking about a pad with decent construction. If you make one by hand, with leaded resistors, and space-wired, with long leads, you will see a big reflection spike at its input. That will still show up on a TDR, as spikes, before the end. On a network analyzer, you would still see the 20 dB reductions, but as frequency goes up, the amount of the reflection also goes up. Which is why you would want to use a really good pad, that has been characterized for the frequency range you are working in. For SPDIF, I would use one spec'ed to 100 MHz. A hand-built one won't cut it, but would still be of great benefit."
I'm not 100% sure but it does seem to make sense. So maybe there is something in using a pad, if it is of good enough quality: having less reflection than the non-padded line to begin with. That seems to be an important condition. if that isn't the case it makes matters worse.
jd
AH Jan, finally! And pass on my thanks to Jocko too - I did miss the nuance of the double attenuation
Yes these attenuators are closely speced up to 2GHz operation so just might be of good enough quality & for $12 are you really taking a big gamble
Of course there is no benefit to doing any of this if:
- You can't attenuate the signal to start with
- The transmission line is already well terminated
Yes these attenuators are closely speced up to 2GHz operation so just might be of good enough quality & for $12 are you really taking a big gamble
Of course there is no benefit to doing any of this if:
- You can't attenuate the signal to start with
- The transmission line is already well terminated
Last edited:
Yes, Jan, signals sent through a pad are reduced.
But the reflection is from where, the arrival (terminating) point?? In which case the reflection is still there, and IF the point it arrives back at - the pad - is not a perfect termination then it bounces again... what is the difference between either of the two terminals of the pad and the output of the DAC? Answer that, and maybe something will be revealed?
... the pad it seems to me might damp a situation where there is a reflection traveling THROUGH the attenuator... but reflections are due to mismatches, yes? - but as you asked, Jan, does it really change the ratio of reflection to signal in a mismatched section?
Now, the wild card is IF there happens to be a resonant length that is involved in the mismatch causing reflections, THEN, breaking up the line with a pad NOT on the end(s) will perhaps change the situation - but so would using a different length cable.
I think a whole lot of what happens here will depend on the mechanism of the reflection(s), that being the source or cause, and where that attenuator is placed.
So where does the magic pad go??
_-_-bear
But the reflection is from where, the arrival (terminating) point?? In which case the reflection is still there, and IF the point it arrives back at - the pad - is not a perfect termination then it bounces again... what is the difference between either of the two terminals of the pad and the output of the DAC? Answer that, and maybe something will be revealed?
... the pad it seems to me might damp a situation where there is a reflection traveling THROUGH the attenuator... but reflections are due to mismatches, yes? - but as you asked, Jan, does it really change the ratio of reflection to signal in a mismatched section?
Now, the wild card is IF there happens to be a resonant length that is involved in the mismatch causing reflections, THEN, breaking up the line with a pad NOT on the end(s) will perhaps change the situation - but so would using a different length cable.
I think a whole lot of what happens here will depend on the mechanism of the reflection(s), that being the source or cause, and where that attenuator is placed.
So where does the magic pad go??
_-_-bear
The reflection is generated wherever it encounters an impedance anomoly - the point is the pad is a perfect termination & is speced out to 2GHz - have a look at the datasheet in the link I gave!
I think Jan's quote from Jocko answers this - the pad does double duty on the reflection compared to the signal!
Why not do an experiment - that was the whole point of starting this thread
Last edited:
Jkeny, I think you are not grasping what I am saying. The pad can be perfect, fine, we can accept that for the moment. What is the cause or source of the reflection in the first place??
Jocko & Jan theorized a nice idea. But it fails to adequately explain anything much unless the source of the mismatch is identified first. Where is the mismatch? That tells us where the reflection is. We need to know this, because if the DAC looks like a perfect 75 ohm load and there is still a reflection, putting another 75ohm load at the DAC, in series with the signal will change nothing except the level going into the DAC, eh??
Where do you/they say to place the pad??
Output of the transport?
Input to the DAC?
Middle of two cables??
Why?
IF all you are doing is "correcting" the termination of the input to the DAC, then maybe there is a problem in the design of the DAC that ought to addressed rather than the "band aid" applied, since the band aid might introduce other problems - as previously noted.
What we're talking about is SWR, or standing wave ratio.
No matter what, the reflection in any reasonably terminated set up is going to be very very low, meaning that the reflections are also rather low... otoh if people are using cables that are in fact NOT 75ohm cables, then there might be a mismatch, I guess. But adding a 75ohm perfect termination to the end of a non-75ohm cable IS a mismatch!
Adding a 75 ohm termination to the output of the transport and before a non-75 ohm cable is a mismatch! Afaik that is...
I think the "experimentation" needs to be done by those who are selling these things, they have to state how to properly utilize the device(s)??
The gear needed to actually see what is happening is not inexpensive, and I do not own any TDR stuff that will work out to 2gHz or so... how about you?
_-_-bear
Jocko & Jan theorized a nice idea. But it fails to adequately explain anything much unless the source of the mismatch is identified first. Where is the mismatch? That tells us where the reflection is. We need to know this, because if the DAC looks like a perfect 75 ohm load and there is still a reflection, putting another 75ohm load at the DAC, in series with the signal will change nothing except the level going into the DAC, eh??
Where do you/they say to place the pad??
Output of the transport?
Input to the DAC?
Middle of two cables??
Why?
IF all you are doing is "correcting" the termination of the input to the DAC, then maybe there is a problem in the design of the DAC that ought to addressed rather than the "band aid" applied, since the band aid might introduce other problems - as previously noted.
What we're talking about is SWR, or standing wave ratio.
No matter what, the reflection in any reasonably terminated set up is going to be very very low, meaning that the reflections are also rather low... otoh if people are using cables that are in fact NOT 75ohm cables, then there might be a mismatch, I guess. But adding a 75ohm perfect termination to the end of a non-75ohm cable IS a mismatch!
Adding a 75 ohm termination to the output of the transport and before a non-75 ohm cable is a mismatch! Afaik that is...
I think the "experimentation" needs to be done by those who are selling these things, they have to state how to properly utilize the device(s)??
The gear needed to actually see what is happening is not inexpensive, and I do not own any TDR stuff that will work out to 2gHz or so... how about you?
_-_-bear
Could we all try to keep our eye on the ball here?
Now we're talking about a reflection which is travelling back to the SOURCE.
IF (and only if) the reflection is re-reflected at the source termination (which has its own coefficient of reflection, which may be ZERO) will it be seen at the receiver.
You're not even capable of understanding your own ill-conceived thought experiments.
The problem, jkeny, with keeping an 'open mind', is that somebody will most assuredly come along and put something IN it.
It would be nice to think that there's a crack in everything. That warp drive will be discovered one day. But there isn't. The stuff you learned at school (or didn't) has been argued and chewed over for decades or centuries by people smarter and more highly motivated than even I. Before you can do anything NEW in engineering, you have to be fully cognizant of prior art. And have a healthy respect for it. 'Shoulders of giants', remember?
It's hubris, foolish pride, to imagine that you can come along with a $12 device and improve perceptibly on the performance of a device constructed by experts. Otherwise they'd be giving out attenuators with every SPDIF transmitter sold. Or at least making recommendations that you should buy one in the manual.
Sometimes I really wonder what you tweakers have got instead of common sense.
w
Last edited:
- Status
- Not open for further replies.