Today I went have a look at the Audiokits site:
http://www.audiokits.com/tweaks.asp
I was surprised to find the following claim there:
>
"Significant improvements can be achieved by using terminated cables between pieces of audio gear. This is especially important when the distance between the audio components is several meters. Let's use this as an example:
Normally the output impedance of a pre-amp is in the order of 50 to several hundred ohms. The input impedance of the power amplifier on the other hand is in the tens to hundreds of kilo Ohm. Connecting them together with a shielded cable means that the cable sees a low driving impedance and a very high load impedance. The result is that the signal you are sending from the pre-amp bounces back from the other end resulting in echoes. The signal will be smeared by the time it arrives at the power amp. This is practically independent of whether you are using a cheap shielded cable, a professional coax cable or a fancy $1000.00/m silver cable. The reason is that the cable is not terminated with its characteristic impedance.
If you terminate the cable at BOTH ends with its characteristic impedance, then it appears to be a purely resistive cable and there will be no reflections (echoes) back from the load."
>
This is the first time I read such a thing. Is that true?
As far as I know you should terminate your cables when you interface digital input/outputs, or at least use the right coaxial and connector, but I never heard of "echoes" (probably meant reflections) happening when interfacing say a CD output with a power amp.
Carlos
http://www.audiokits.com/tweaks.asp
I was surprised to find the following claim there:
>
"Significant improvements can be achieved by using terminated cables between pieces of audio gear. This is especially important when the distance between the audio components is several meters. Let's use this as an example:
Normally the output impedance of a pre-amp is in the order of 50 to several hundred ohms. The input impedance of the power amplifier on the other hand is in the tens to hundreds of kilo Ohm. Connecting them together with a shielded cable means that the cable sees a low driving impedance and a very high load impedance. The result is that the signal you are sending from the pre-amp bounces back from the other end resulting in echoes. The signal will be smeared by the time it arrives at the power amp. This is practically independent of whether you are using a cheap shielded cable, a professional coax cable or a fancy $1000.00/m silver cable. The reason is that the cable is not terminated with its characteristic impedance.
If you terminate the cable at BOTH ends with its characteristic impedance, then it appears to be a purely resistive cable and there will be no reflections (echoes) back from the load."
>
This is the first time I read such a thing. Is that true?
As far as I know you should terminate your cables when you interface digital input/outputs, or at least use the right coaxial and connector, but I never heard of "echoes" (probably meant reflections) happening when interfacing say a CD output with a power amp.
Carlos
This with terminated cables are very important when the wave length is near or shorter than the cable length. This is a technical fact.
The wave length of 20 kHz is 15 km.
You can make up your own mind when you think that this has minor importance. Is for 1 km cable, 100 meters, 10 meters, 1 meter????
But, it's cool to have impedance matched connections. It's not hard at all to measure the charateristc impedance of a cable. Take a known length, 10-50 meters, at least 10 meters. Then you must have a oscilloscope which can read 100 ns without a problem. You must also have a signal source (TTL) which can give you sharp square waves. Leave the cable open and then add a potentiometer or a resistor. Measure only in the beginning of the cable. When the reflection is at minimum you have the impedance. Audio cables have often impedances down to 25 ohms. You can see this by your eyes. If the isolation is thin between the shield and the inner cable, then the cable has low impedance. Compare a RG-59 coax cable!
The wave length of 20 kHz is 15 km.
You can make up your own mind when you think that this has minor importance. Is for 1 km cable, 100 meters, 10 meters, 1 meter????
But, it's cool to have impedance matched connections. It's not hard at all to measure the charateristc impedance of a cable. Take a known length, 10-50 meters, at least 10 meters. Then you must have a oscilloscope which can read 100 ns without a problem. You must also have a signal source (TTL) which can give you sharp square waves. Leave the cable open and then add a potentiometer or a resistor. Measure only in the beginning of the cable. When the reflection is at minimum you have the impedance. Audio cables have often impedances down to 25 ohms. You can see this by your eyes. If the isolation is thin between the shield and the inner cable, then the cable has low impedance. Compare a RG-59 coax cable!
terminated cables
The point of terminating cables to thier characteristics impedance is for stability of the circuit driving the cable and for RF rejection.
Read the ap notes or data sheet for just about any high speed op amp you will see a section on driving capacitive loads. (such as an unterminated cable for example) I use a RC filter with the cables characteristic impedance as the resitive element and a capacitor of a few hundred picofarads. This terminatesthe cable and filters the amp at RF frequencies but is still easy the drive at audio frequencies. A build out resistor at the driven end of the cable is a good idea also. In a world of cell phones and microprocessors you had better treat every circuit as a RF circuit. RF will find its way into your audio circuits. Shame on you Peranders for talking about this in reference to 20 KHz. This is a nobrainer that any decent audio engineer knows.
http://www.elantec.com/pages/apppdf/d40954.pdf
http://www.e-insite.net/ednmag/archives/1996/042596/09df3.htm
http://www.analog.com/UploadedFiles/Application_Notes/180434159AN257.pdf
The point of terminating cables to thier characteristics impedance is for stability of the circuit driving the cable and for RF rejection.
Read the ap notes or data sheet for just about any high speed op amp you will see a section on driving capacitive loads. (such as an unterminated cable for example) I use a RC filter with the cables characteristic impedance as the resitive element and a capacitor of a few hundred picofarads. This terminatesthe cable and filters the amp at RF frequencies but is still easy the drive at audio frequencies. A build out resistor at the driven end of the cable is a good idea also. In a world of cell phones and microprocessors you had better treat every circuit as a RF circuit. RF will find its way into your audio circuits. Shame on you Peranders for talking about this in reference to 20 KHz. This is a nobrainer that any decent audio engineer knows.
http://www.elantec.com/pages/apppdf/d40954.pdf
http://www.e-insite.net/ednmag/archives/1996/042596/09df3.htm
http://www.analog.com/UploadedFiles/Application_Notes/180434159AN257.pdf
In broadcast studio environments we used to use 600ohm source - 600ohm load.
This was because the technology was borrowed from the telecom industry, where long lines happen.
That system was abandoned years ago, for Low Z source - high Z load.
The main signal modifying component is the cable's capacitance, which is not significant when the source Z is low.
Probably to the horror of budding audiofiles:
Full bandwidth analogue signals are still sent down bundled twisted pair cables, sometimes for hundreds of metres. As long as the signal is balanced, and the pair has a twist, the losses and crosstalk are acceptable.
To add insult to injury:
AES/EBU Digital signals are also sent down twisted pairs, but in this case we obey the matching rule: 110ohm source - 110ohm load, because the frequency is higher.
Recently there has been a move to use CAT5 (Ethernet) cable for less critical digital applications as it is so cheap. So far the results have shown it is superior to some "digital" cable, dispite the slight mis-match.
These signals are the balanced counterpart on SPDIF, differing only in level (5v) and pathology.
This was because the technology was borrowed from the telecom industry, where long lines happen.
That system was abandoned years ago, for Low Z source - high Z load.
The main signal modifying component is the cable's capacitance, which is not significant when the source Z is low.
Probably to the horror of budding audiofiles:
Full bandwidth analogue signals are still sent down bundled twisted pair cables, sometimes for hundreds of metres. As long as the signal is balanced, and the pair has a twist, the losses and crosstalk are acceptable.
To add insult to injury:
AES/EBU Digital signals are also sent down twisted pairs, but in this case we obey the matching rule: 110ohm source - 110ohm load, because the frequency is higher.
Recently there has been a move to use CAT5 (Ethernet) cable for less critical digital applications as it is so cheap. So far the results have shown it is superior to some "digital" cable, dispite the slight mis-match.
These signals are the balanced counterpart on SPDIF, differing only in level (5v) and pathology.
Re: terminated cables
Using a cap for RF rejection is certainly a point worth considering. Many will claim it's not, but at some point you will need to protect yourself from RF.
Most low impedance outputs (like those on CDP) are also series protected with 50 to 200 ohm resistors.
Also most amps have an input resistor, going from tenth Ks to a few hundred Ks. An input cap is also there usually.
So why should we add a resistor at both ends, as the article suggests? Just in case?
Carlos
Fred Dieckmann said:
Using a cap for RF rejection is certainly a point worth considering. Many will claim it's not, but at some point you will need to protect yourself from RF.
Most low impedance outputs (like those on CDP) are also series protected with 50 to 200 ohm resistors.
Also most amps have an input resistor, going from tenth Ks to a few hundred Ks. An input cap is also there usually.
So why should we add a resistor at both ends, as the article suggests? Just in case?
Carlos
Well, I am a little concerned about adding source and load termination resistors. For one, and this is less true of DIY'ers, we may not know what the output impedance (Zout) of the source is, or what the load impedance (Zin) of the load is. Typical Zin's seem to be in 30K to 47K range for power amps. For preamps, especially solid state, the Zouts can be pretty low.
But let's assume you use 50 ohm coax as your interconnect cable. If the Zout is around 50 ohms, then you don't need any source termination. If it is much lower, then you would need a series resistor. At the amplifier end, you would have to add a 50 ohm resistor to ground. Well, if we do this, we have a matched system, but our available voltage from the preamp has just been cut in half, and the power requirements from the preamp are now up to 1,000 times greater than it was (not that most preamps could not drive a 50 ohm load).
If the intent of all of this is to prevent wave reflections of audio signals in the cable, I don't see how it can help. As has been said, audio wavelengths are very long, and even if we assume that multiple harmonics of 20 KHz are audible, standing waves effects for a 2 meter cable won't come into play until we get up into the AM broadcast band.
We try to terminate high-frequency op amps as they have high open and closed loop bandwidths. Most audio amps have low pass filters at their inputs to keep out RF, and they have their closed loop gains rolled off to well under 100 KHz. In high frequency op amp design and high speed digital design we need to take precauitions to prevent reflections, but for our low frequency audio amplifiers, it is not a concern. While RF can find its way in, it will generally be via radiation from a nearby RF transmitter. Here, we are not so interested in providing a terminated load for it to prevent reflections as we are in simply attenuating it as much as possible - hence our low pass filtering. In other words, if there are reflections on our amplifier input cable from a nearby radio transmitter, we couldn't care less if the RF signal gets distorted or not due to reflections - we simply want to attenuate it and ensure that our amplifer does not respond to it.
As always, there may be some mysterious way that providing for termination helps with audio, but I have to think the mechanism is not due to the prevention of standing waves or reflections, but something else. I remain highly skeptical.
But let's assume you use 50 ohm coax as your interconnect cable. If the Zout is around 50 ohms, then you don't need any source termination. If it is much lower, then you would need a series resistor. At the amplifier end, you would have to add a 50 ohm resistor to ground. Well, if we do this, we have a matched system, but our available voltage from the preamp has just been cut in half, and the power requirements from the preamp are now up to 1,000 times greater than it was (not that most preamps could not drive a 50 ohm load).
If the intent of all of this is to prevent wave reflections of audio signals in the cable, I don't see how it can help. As has been said, audio wavelengths are very long, and even if we assume that multiple harmonics of 20 KHz are audible, standing waves effects for a 2 meter cable won't come into play until we get up into the AM broadcast band.
We try to terminate high-frequency op amps as they have high open and closed loop bandwidths. Most audio amps have low pass filters at their inputs to keep out RF, and they have their closed loop gains rolled off to well under 100 KHz. In high frequency op amp design and high speed digital design we need to take precauitions to prevent reflections, but for our low frequency audio amplifiers, it is not a concern. While RF can find its way in, it will generally be via radiation from a nearby RF transmitter. Here, we are not so interested in providing a terminated load for it to prevent reflections as we are in simply attenuating it as much as possible - hence our low pass filtering. In other words, if there are reflections on our amplifier input cable from a nearby radio transmitter, we couldn't care less if the RF signal gets distorted or not due to reflections - we simply want to attenuate it and ensure that our amplifer does not respond to it.
As always, there may be some mysterious way that providing for termination helps with audio, but I have to think the mechanism is not due to the prevention of standing waves or reflections, but something else. I remain highly skeptical.
dhaen said:
Fast birates demands correct impedances and AES/EBU requires cables with 110 ohm impedance. More important with faster bit rates....
600 ohms impdance came from the time when we wanted to get maximium power out of a system which accurs when the load impedance is the same as the source impedance.
IMO That was not the main reason.
The main reason is that if the cable has a characteristic imedance of 600 ohms, then there is no loss as a result of the cable capacitance or inductance. In fact you can represent the cable as an infinite series of LC circuits. (Check Transmission line theory). Of course after 10Km there are losses, but these are to do with dielectric, crosstalk, and discontinuity mismatch.
For a matched system to work all three:
The source,
The cable,
The load.
Must be of the same impedance.
But it makes no sense at audio frequencies.
Just adding resistors across the conductors is nonsence.
Re: terminated cables
Er... pardon? I thought that the use of cables having a given characteristic impedance was so that the source 'saw' the same impedance at the input of the cable equal to the load...... The point being that at RF, the maximum power is transferred when the source impedance equals the load impedance.
er... how does it terminate the cable? In order for the cable to be terminated, you need the characteristic impedance at both ends, not just at one end. After all, the definition of characteristic impedance is "the impedance seen looking into a cable when the cable is terminated in that impedance".
Accepted... to some extent. The resistor's being the same value as the cable's impedance seems to me irrelevant though. However, at audio frequencies a piece of coaxial cable will be primarily capacitive. In many systems it is, in fact, modelled as such (a capacitor).
Why? And how do you define "build out resistor"? To me a build out resistor can either be in series with a circuit or in parallel. Which? And why? And please don't say "It sounds better". Come up with a valid reason that can be checked.
I had a look at your references. The first two talked about making high frequency op amps stable and matching (not audio op amps) and the third just talked about making op amps stable. Hmmmmm.
Here we do agree.
Why? I think he's right. And 20kHz is a good figure to use as it has a short wavelength for audio. Any lower frequency and the wavelength is even longer.
I think also you'd better go and look at the specs of many cables. The characteristic impedance is given for a certain frequency band. For example, I've worked with certain cables that have a characteristic impedance of 50 ohms. Fair enough, but its only above 1MHz. Below that the impedance drops away dramatically. Some cables are specified above 10kHz, others above 10MHz.
Compliments of the season to you and your's.
Cheers, Keith
Fred Dieckmann said:
Er... pardon? I thought that the use of cables having a given characteristic impedance was so that the source 'saw' the same impedance at the input of the cable equal to the load...... The point being that at RF, the maximum power is transferred when the source impedance equals the load impedance.
er... how does it terminate the cable? In order for the cable to be terminated, you need the characteristic impedance at both ends, not just at one end. After all, the definition of characteristic impedance is "the impedance seen looking into a cable when the cable is terminated in that impedance".
Accepted... to some extent. The resistor's being the same value as the cable's impedance seems to me irrelevant though. However, at audio frequencies a piece of coaxial cable will be primarily capacitive. In many systems it is, in fact, modelled as such (a capacitor).
Why? And how do you define "build out resistor"? To me a build out resistor can either be in series with a circuit or in parallel. Which? And why? And please don't say "It sounds better". Come up with a valid reason that can be checked.
I had a look at your references. The first two talked about making high frequency op amps stable and matching (not audio op amps) and the third just talked about making op amps stable. Hmmmmm.
Here we do agree.
Why? I think he's right. And 20kHz is a good figure to use as it has a short wavelength for audio. Any lower frequency and the wavelength is even longer.
I think also you'd better go and look at the specs of many cables. The characteristic impedance is given for a certain frequency band. For example, I've worked with certain cables that have a characteristic impedance of 50 ohms. Fair enough, but its only above 1MHz. Below that the impedance drops away dramatically. Some cables are specified above 10kHz, others above 10MHz.
Compliments of the season to you and your's.
Cheers, Keith
It's a known fact that the speed (characteristic impedance) of a cable varies with fr
Not by much........ Dielectric losses and skin effect at very high frequencies (10s to 100s of mHz) but pretty irrelevent for most video, RF, and digital signals. Build out resistor reffered to is series resistor at source. Cables and PCB traces are often terminated at one end only. This particularly true for Digital design. Cables need to be terminated at RF frequencies as line lengths for audio freqencies are too short to be considered as transmission lines.
http://www.chipcenter.com/oltu/netsim/si/termination/
http://www.sigcon.com/Pubs/edn/TransmissionLine.htm
http://www.sigcon.com/Pubs/edn/LossyLine.htm
Not by much........ Dielectric losses and skin effect at very high frequencies (10s to 100s of mHz) but pretty irrelevent for most video, RF, and digital signals. Build out resistor reffered to is series resistor at source. Cables and PCB traces are often terminated at one end only. This particularly true for Digital design. Cables need to be terminated at RF frequencies as line lengths for audio freqencies are too short to be considered as transmission lines.
http://www.chipcenter.com/oltu/netsim/si/termination/
http://www.sigcon.com/Pubs/edn/TransmissionLine.htm
http://www.sigcon.com/Pubs/edn/LossyLine.htm
The lack of a need of a terminating resistor must be a great relief to those on the forum that argue that the colour of a connecting wire insulation affects the sound. I mean to say, those different coloured stripes on the terminating resistor are going to play havoc with the musicality of the wire, are they not? I suppose we could get around it by using a different numbering system than base 10, therefore using a more coherent set of colours for the resistor values...
The RCA handbook has many amplifier designs with performance specs.
The average input impedance of these designs is about 22K.
Signal to noise ratio measurements taken with a ten foot cable terminated at the amplifier end by 2K2 show a 10dB improvement.
I can hear 10dB.
I would not go less than 2K2 as the dynamics of most pre-amps will suffer. Just becuse your gear is rated to drive 600R doesn't mean it will sound the best doing so.
The average input impedance of these designs is about 22K.
Signal to noise ratio measurements taken with a ten foot cable terminated at the amplifier end by 2K2 show a 10dB improvement.
I can hear 10dB.
I would not go less than 2K2 as the dynamics of most pre-amps will suffer. Just becuse your gear is rated to drive 600R doesn't mean it will sound the best doing so.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.