When you have a MS in EE and have designed digital converters in the 20ghz range for Princeton Us High Energy Particle Physics dept, then you are entitled to laugh. I will bet money you have never heard his DAC either. I have and I have forty years of classical music recording experience, using Schoeps, Studer, Stellavox, RTW, etd. I think I know something about sound reproduction.
@john
you don't need a MS in EE to understand simple sentence:
Lowest interwinding capacitance = highest leakage inductance
you don't need a MS in EE to understand simple sentence:
Lowest interwinding capacitance = highest leakage inductance
Have you read the white papers on the Scientific Conversion website? These might help explain the rational for the choice of low capacitance. Regards
I am oldfashioned guy and prefer to breadboard and measure, not to read and simulate.
They are wrong, low interwinding capacitance is wrong way to follow.
They are wrong, low interwinding capacitance is wrong way to follow.
Last edited:
Just one nitpick in a post I otherwise broadly agree with. In the case of electrostatic screens in power transformers, competent grounding is beyond the CD player designer's ability to ensure. If the consumer doesn't provide a clean earth then 'earthed' electrostatic screens on mains transformers can turn from being a blessing into a curse.
Do please explain why leakage inductance is more of an issue than interwinding capacitance. This enquiring mind wants to know 🙂
you can Google for it, you can read Jocko's posts, or you can terminate transmission line with inductor to see effect with your own eyes.
Choice is yours 🙂
Choice is yours 🙂
I'm already aware of Jocko's posts, haven't seen anything about the tradeoffs. I'm aware of what happens when a transmission line is terminated with an inductor too. So that's your only reason for saying 'they are wrong' ?
Hi,
However, those who observe that "More Programmes = More Jitter" do not make such a suggestion, they simply do not evaluate the mechanism and merely observe.
So their observations are entirely valid, yet you find it necessary to pour scorn onto them, may I enquire as to the why?
Depends on the amounts of Jitter.
There are many ways, but they all require of course systems whose inherent jitter is low. Some Computer sound cards have reasonably low jitter, but they are very few. Their nature of having the clock drive D to A and A to D will hide this jitter from a loopback measurement, even with SPDIF output and analogue in into the soundcard.
When using two of the same cards "crosscoupled" one can see the jitter quite clearly and plainly though, a few nanoseconds are not unusual, making them mostly useless for measuring jitter.
An AP2 is useful up to a point, for really low jitter it is not suitable. Analysers for Clock Phasenoise (which is another way of saying jitter) do exist that allow very levels of clock jitter to be measured.
The guy with the handle JosephK uses, I believe a very nice (and expensive - it costs more than most peoples cars are worth) LeCroy digital 'scope.
Ciao T
However, those who observe that "More Programmes = More Jitter" do not make such a suggestion, they simply do not evaluate the mechanism and merely observe.
So their observations are entirely valid, yet you find it necessary to pour scorn onto them, may I enquire as to the why?
Depends on the amounts of Jitter.
There are many ways, but they all require of course systems whose inherent jitter is low. Some Computer sound cards have reasonably low jitter, but they are very few. Their nature of having the clock drive D to A and A to D will hide this jitter from a loopback measurement, even with SPDIF output and analogue in into the soundcard.
When using two of the same cards "crosscoupled" one can see the jitter quite clearly and plainly though, a few nanoseconds are not unusual, making them mostly useless for measuring jitter.
An AP2 is useful up to a point, for really low jitter it is not suitable. Analysers for Clock Phasenoise (which is another way of saying jitter) do exist that allow very levels of clock jitter to be measured.
The guy with the handle JosephK uses, I believe a very nice (and expensive - it costs more than most peoples cars are worth) LeCroy digital 'scope.
Ciao T
Hi,
However, those who observe that "More Programmes = More Jitter" do not make such a suggestion, they simply do not evaluate the mechanism and merely observe.
So their observations are entirely valid, yet you find it necessary to pour scorn onto them, may I enquire as to the why?
BTW, given the multitasking nature of PC's and depending on soundcards design etc., it is certainly also possible to have additional sources of jitter which result from the CPU returning to service an Audio Stream a few clock cycles early or late, these may be buffered out, but they may not be...
Depends on the amounts of Jitter.
There are many ways, but they all require of course systems whose inherent jitter is low. Some Computer sound cards have reasonably low jitter, but they are very few. Their nature of having the clock drive D to A and A to D will hide this jitter from a loopback measurement, even with SPDIF output and analogue in into the soundcard.
When using two of the same cards "crosscoupled" one can see the jitter quite clearly and plainly though, a few nanoseconds are not unusual, making them mostly useless for measuring jitter.
An AP2 is useful up to a point, for really low jitter it is not suitable. Analysers for Clock Phasenoise (which is another way of saying jitter) do exist that allow very levels of clock jitter to be measured.
The guy with the handle JosephK uses, I believe a very nice (and expensive - it costs more than most peoples cars are worth) LeCroy digital 'scope that seems nice in measuring jitter from many different angles, I wants one.
Ciao T
However, those who observe that "More Programmes = More Jitter" do not make such a suggestion, they simply do not evaluate the mechanism and merely observe.
So their observations are entirely valid, yet you find it necessary to pour scorn onto them, may I enquire as to the why?
BTW, given the multitasking nature of PC's and depending on soundcards design etc., it is certainly also possible to have additional sources of jitter which result from the CPU returning to service an Audio Stream a few clock cycles early or late, these may be buffered out, but they may not be...
Depends on the amounts of Jitter.
There are many ways, but they all require of course systems whose inherent jitter is low. Some Computer sound cards have reasonably low jitter, but they are very few. Their nature of having the clock drive D to A and A to D will hide this jitter from a loopback measurement, even with SPDIF output and analogue in into the soundcard.
When using two of the same cards "crosscoupled" one can see the jitter quite clearly and plainly though, a few nanoseconds are not unusual, making them mostly useless for measuring jitter.
An AP2 is useful up to a point, for really low jitter it is not suitable. Analysers for Clock Phasenoise (which is another way of saying jitter) do exist that allow very levels of clock jitter to be measured.
The guy with the handle JosephK uses, I believe a very nice (and expensive - it costs more than most peoples cars are worth) LeCroy digital 'scope that seems nice in measuring jitter from many different angles, I wants one.
Ciao T
@abraxalito
why there should be tradeoffs? I just observe the waveform.
Transformers with high leakage inductance have overshots and spikes.
HF energy have to go somewhere and since it is not coupled from primary to secondary (therefore name "leakage inductance") it will create overshots and spikes. Those overshots and spikes will mess with rise & fall times thus introducing jitter.
You can try to cancel those spikes with Zobel, but it takes time to play with different capacitor values (few pF's) and results are not always satisfactory. It is much easier to start with low leakage transformer.
If common mode noise is your primary concern, fast video opamps and line drivers are right answer. With low source resistances and for low gains (G=+2) there are CFA's with over 1 GHz BW and several kV/µs slew rate
why there should be tradeoffs? I just observe the waveform.
Transformers with high leakage inductance have overshots and spikes.
HF energy have to go somewhere and since it is not coupled from primary to secondary (therefore name "leakage inductance") it will create overshots and spikes. Those overshots and spikes will mess with rise & fall times thus introducing jitter.
You can try to cancel those spikes with Zobel, but it takes time to play with different capacitor values (few pF's) and results are not always satisfactory. It is much easier to start with low leakage transformer.
If common mode noise is your primary concern, fast video opamps and line drivers are right answer. With low source resistances and for low gains (G=+2) there are CFA's with over 1 GHz BW and several kV/µs slew rate
Dunno, I didn't design the laws of physics. I just know that lower inter-winding cap means higher leakage inductance and vice versa.
The leakage inductance is in series with the primary right? So did you terminate prior to the transformer or after it when you looked at the waveforms? And what makes overshoots and spikes a particular problem? Did your receiver object in some way? What's the difference in jitter you observed between low cap and high cap transformers?
CM noise is a major concern. Please explain how these fix that coz I can't see it myself?
Yeah, I'm aware of various of these.
This was not the point of my post, but if you wish to view it like that then so be it.
Hi,
I view it as written. You wrote above:
And the answer is, yes, by their presence they will introduce jitter on typical hardware platforms. And they will do it from a data-streaming viewpoint, as the data cannot be streamed without clock (which you indirectly allow for, just as if that was something totally different).
Ciao T
I view it as written. You wrote above:
And the answer is, yes, by their presence they will introduce jitter on typical hardware platforms. And they will do it from a data-streaming viewpoint, as the data cannot be streamed without clock (which you indirectly allow for, just as if that was something totally different).
Ciao T
Engineering seems to always have trade-offs. If you minimize leakage inductance at the expense of isolation capacitance then you've defeated the reason for the transformers in the first place and simply added an error stage.
Seems to me there is a region of compromise where the isolation minimizes the influence of common mode noise on the received clocks transitions and where the leakage inductance is small to where its presence does not significantly degrade the secondary response or cause reflections on the primary to influence the next edge.
My 2 cents
-Antonio
Seems to me there is a region of compromise where the isolation minimizes the influence of common mode noise on the received clocks transitions and where the leakage inductance is small to where its presence does not significantly degrade the secondary response or cause reflections on the primary to influence the next edge.
My 2 cents
-Antonio
Well it turns out your thinking is mistaken 😀
First link might be able to teach me something except for the minor detail that without a login, the schematics are not displayed.
The second link I agree is great work by JosephK (I was following it almost in real time) but doesn't address the question I put to you about how jitter varies between the two extremes of transformer design.
Thirdly, CMRR of an opamp doesn't do anything to help reduce common mode noise in the system, it merely attenuates it in the signal. So that's a red herring.
Zero out of three - room for improvement wouldn't you say? 😛
<edit> Sorry, I missed a link and its a good one - so make that one out of four.
Last edited:
My sentiments entirely. Engineering is the art of optimisation and that's almost never done at one or other of the extremes.
Yes, although I'm more concerned about the influence of CM noise on the system as a whole. What goes in via the SPDIF will come out via the audio feeds. Having higher leakage inductance means a higher impedance to that noise is presented since the inductance is in series with the noise. Reflections have to be very bad to corrupt the data - if I've got great jitter rejection further down the datapath (like a secondary PLL, digitally implemented), I find myself wondering why they would need to concern me unduly.
- Status
- Not open for further replies.