Size matters
Many of the 1/2 meter digital cables I have heard sound much worse than the 1 meter version. The reflection time for that length of cable can put the refected signal back on the rising edge on the digital signal. This depends on the rise of the cable driver circuit and the velocity factor of the coax. I have found 1.5 meters to sound better than one meter cable and that was the standard length of the cables I designed for Audient Technologies.
Many of the 1/2 meter digital cables I have heard sound much worse than the 1 meter version. The reflection time for that length of cable can put the refected signal back on the rising edge on the digital signal. This depends on the rise of the cable driver circuit and the velocity factor of the coax. I have found 1.5 meters to sound better than one meter cable and that was the standard length of the cables I designed for Audient Technologies.
"I shouldn't interfere here but is reflection really something to care about?"
Only if you actually care what it sounds like. It is pretty audible in my experience. I believe Jock did some experimants with reducing the cable length incrementally and listening to the results. He said it was possible to find a length that sounded really bad.
How much do you really get?
Enough to be easily measureable on a 100MHz scope. Obviously it is a function of the impedance discontinuity, signal rise time, cable velocity factor, length, and lossiness. he likes loooong cables, like 10 feet and more. Lossy cables also often work well.
This is a constant defect and do you really get distorted signals, after the reciever?
Reflections cause noise. If timing is such that the reflection returns at the part rising edge where the the receiver is looking for a logic transistion, jitter can occur. The right reflection time can also result in undershoot or increased ringing. They can also contribute to EMI and the effects of that on sonics can be pretty bad. Don't foget that these reflections bounce back from the source end to the reciever. The presence of EMI, logic power supply noise, and receiver hysteresis compond this problem.
1 meter is 5 ns in one direction
That depends on the velocity factor of the cable and is a function of the dielectric constant of the cable insulator. Interestingly digital cable are often directional. Differences in termination and the fact that the cables center conductors distance from the sheild result in changes in the characteristic impedance down the cable length. Cable direction change the time delays of the reflections caused by these impedance continuities. One of the audio magazines reported different jitter measurements from changing the cable direction. Anybody still doubt digital cables can sound different.
For a good Signal integrity book:
http://www.sigcon.com/books.htm
I went to one of his seminars also. He invented the field of study of signal integrity which is crucial in high speed logic design.
Only if you actually care what it sounds like. It is pretty audible in my experience. I believe Jock did some experimants with reducing the cable length incrementally and listening to the results. He said it was possible to find a length that sounded really bad.
How much do you really get?
Enough to be easily measureable on a 100MHz scope. Obviously it is a function of the impedance discontinuity, signal rise time, cable velocity factor, length, and lossiness. he likes loooong cables, like 10 feet and more. Lossy cables also often work well.
This is a constant defect and do you really get distorted signals, after the reciever?
Reflections cause noise. If timing is such that the reflection returns at the part rising edge where the the receiver is looking for a logic transistion, jitter can occur. The right reflection time can also result in undershoot or increased ringing. They can also contribute to EMI and the effects of that on sonics can be pretty bad. Don't foget that these reflections bounce back from the source end to the reciever. The presence of EMI, logic power supply noise, and receiver hysteresis compond this problem.
1 meter is 5 ns in one direction
That depends on the velocity factor of the cable and is a function of the dielectric constant of the cable insulator. Interestingly digital cable are often directional. Differences in termination and the fact that the cables center conductors distance from the sheild result in changes in the characteristic impedance down the cable length. Cable direction change the time delays of the reflections caused by these impedance continuities. One of the audio magazines reported different jitter measurements from changing the cable direction. Anybody still doubt digital cables can sound different.
For a good Signal integrity book:
http://www.sigcon.com/books.htm
I went to one of his seminars also. He invented the field of study of signal integrity which is crucial in high speed logic design.
Fred Dieckmann said:
Thanks, kind of what I thought. I know all to well what reflections can do after working with networking for a fair amount of time and learning quickly the difference between proper and bad termination.
As a reflection (!), did you, or somebody else that you know of, ever test to use a much longer cable than needed and put connectors at reasonable distance, say less than 1 m, but extending the line some extra meters from both end points and terminating there. I just thought that maybe one could get the reflection attenuated slightly and moving out in time to where it could be less harmful. Stupid or interesting?
Attachments
tiroth said:
I am not sure that is how it works but I know that both ends for a 50 Ohm network need to be terminated with 50 in this configuration. Also I suspect that a reflection is not only a one-off thing but it will reflect at the receiver end back to the sender end and then reflect back again until the reflection has been completely attenuated. By moving the timing out and keeping the connections close (or rather at a perfect distinance) I thought soemthing good might come out of it. Just a thought as usual...
If you have an ideal transimission line properly terminated at the far end, there will be no reflection. The question becomes how far from ideal is the transmission line in question?
Let's say you have an S/PDIF driver wired with some traces to a 75Ω BNC connectors, then a 75Ω BNC cable to another 75Ω BNC connector with a 75Ω terminating resistor and some traces to a receiver. Where does that depart from the ideal? Each connector introduces an impedance discontinuity, from which the signal will reflect. When these reflections reach the transmitter, they are reflected again back to the receiver. This second-order reflection, and all higher even-order reflections, are the ones that can mess up your signal.
The usual way to deal with a problematic second-order reflection is to terminated at both ends, which will attenuate the reflection from the source. The side effect is halving the signal amplitude, but often this doesn't matter.
So, why does cable length matter? The cable length determines when you will see that second-order reflection. If you think of the cable as a fraction of the wavelength of your signal (determined by the risetime, not the frequency), you can determine when you will see the reflection. Taking the example of a signal with a 15ns rise time (i'm guessing here, what is it for SPDIF?), the wavelength on a coaxial cable will be over 7m. Therefore if you have a 0.5m cable, any reflections will be added to the edge that produced them, causing overshoot. This will not flip bits, but it can artificially increase the rise time, leading to jitter.
Take the other extreme. Suppose you have a 30m cable. The second-order reflection will be added at some random place in the signal, probably four or five pulses after the pulse that created it. The poblem with jitter is still there, and a new signal integrity problem may arise. You won't need to worry about it if the reflections are small.
I think if you are using high-quality cables and connectors and you have proper terminations, your attention should be spent on the signal integrity inside your circuit. I doubt that most designs have much improvement to make in the external cabling. And I also fully believe standard short cable lengths are perfectly good.
Let's say you have an S/PDIF driver wired with some traces to a 75Ω BNC connectors, then a 75Ω BNC cable to another 75Ω BNC connector with a 75Ω terminating resistor and some traces to a receiver. Where does that depart from the ideal? Each connector introduces an impedance discontinuity, from which the signal will reflect. When these reflections reach the transmitter, they are reflected again back to the receiver. This second-order reflection, and all higher even-order reflections, are the ones that can mess up your signal.
The usual way to deal with a problematic second-order reflection is to terminated at both ends, which will attenuate the reflection from the source. The side effect is halving the signal amplitude, but often this doesn't matter.
So, why does cable length matter? The cable length determines when you will see that second-order reflection. If you think of the cable as a fraction of the wavelength of your signal (determined by the risetime, not the frequency), you can determine when you will see the reflection. Taking the example of a signal with a 15ns rise time (i'm guessing here, what is it for SPDIF?), the wavelength on a coaxial cable will be over 7m. Therefore if you have a 0.5m cable, any reflections will be added to the edge that produced them, causing overshoot. This will not flip bits, but it can artificially increase the rise time, leading to jitter.
Take the other extreme. Suppose you have a 30m cable. The second-order reflection will be added at some random place in the signal, probably four or five pulses after the pulse that created it. The poblem with jitter is still there, and a new signal integrity problem may arise. You won't need to worry about it if the reflections are small.
I think if you are using high-quality cables and connectors and you have proper terminations, your attention should be spent on the signal integrity inside your circuit. I doubt that most designs have much improvement to make in the external cabling. And I also fully believe standard short cable lengths are perfectly good.
Well, the "standard" AES/EBU balanced cable is a terrible design. It just shouldn't be used.
If you hear differences between different coax cables of the same length, I would have to suggest that some of them are of higher quality than others. Cable quality <em>does</em> make a difference. Defects in the dielectric, poor attachment of the connector, poor quality conductors, and other defects can cause problems. Also some cables aren't really the impedance they claim.
However you will note that I said "If you are using <b>high-quality</b> cables,...", emphasis added.
If you hear differences between different coax cables of the same length, I would have to suggest that some of them are of higher quality than others. Cable quality <em>does</em> make a difference. Defects in the dielectric, poor attachment of the connector, poor quality conductors, and other defects can cause problems. Also some cables aren't really the impedance they claim.
However you will note that I said "If you are using <b>high-quality</b> cables,...", emphasis added.
Pro equipment includes the AES/EBU connection to be compatible with other, existing equipment. But BNC-terminated coax or twinax is clearly superior. Have you ever seen a 4GHz oscilloscope, or a logic analyzer, or a spectrum analyzer, or for that matter a radio, with an AES input? Of course not. it's a cable technology suited to connecting microphones, not fast digital equipment.
Really, you don't have to take my word for it. Hook up your ≥ 100MHz oscilloscope and take a look at the eye pattern.
Really, you don't have to take my word for it. Hook up your ≥ 100MHz oscilloscope and take a look at the eye pattern.
dear all,
Some remarks, I like this discussion and appreciate the contributions
RCA plugs are terrible, they exhibit impedance of 30 ohm which does not match 75 ohm
BNC connectors often show lower transfer impedance, so potentially lower induced noise voltages in your signal
Bandwidth limmiting at the sending end to "just what is required" prevents sending RF energy on the cable and as such limits reflections, in the case an impedance mismatch is made
I myself make use of all above described, and do not hear differences between 1 meter or less or more
Ofcourse the jitter immunity of the DAC play a big role here
all the best
Some remarks, I like this discussion and appreciate the contributions
RCA plugs are terrible, they exhibit impedance of 30 ohm which does not match 75 ohm
BNC connectors often show lower transfer impedance, so potentially lower induced noise voltages in your signal
Bandwidth limmiting at the sending end to "just what is required" prevents sending RF energy on the cable and as such limits reflections, in the case an impedance mismatch is made
I myself make use of all above described, and do not hear differences between 1 meter or less or more
Ofcourse the jitter immunity of the DAC play a big role here
all the best
Guido Tent said:
The impedance discontinuity of an RCA is for a rather short propagation time. A simple network will do a very good bit to null this discontinuity and limit the bandwidth for reflections and noise. Many SPDIF interfaces are to fast. Go drive them with AC cmos logic and listen if you doubt it. I have heard cables under a meter sound real bad, even terminated with good resistive terminations and 75 ohm BNCs.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.