Post by KeithL on Jan 14, 2015 12:44:49 GMT -5
There are a lot of misunderstandings going around about feedback, output impedance, and minimum load.
About Feedback.
It is true that usually, when you see very low distortion specs, they are achieved by using a lot of feedback - the reason being that feedback is a very effective way of reducing distortion. However, most of the misconceptions about how feedback relates to sound quality are based on largely outdated assumptions. There is nothing INHERENTLY bad about applying feedback. Feedback is only detrimental to sound quality when it is used IMPROPERLY. Unfortunately, it's easy to use feedback to make a poorly designed circuit "measure" better, which is why historically many circuits that had excellent specifications, but still managed to sound bad, turned out to have been designed with very high levels of feedback. (It was simply a combination of not understanding how other circuit characteristics affect the result of applying feedback, and of being unwilling or unable to avoid the temptation of "fixing the numbers" with insufficient regard for sound quality. Many of the problems experienced by early designs, like transient distortion, are relatively easy to detect and avoid - once you know what to look for. The original Dynaco PAT4 preamp was the poster child for a solid state preamp that "measured well but sounded awful".) Rather than claim "feedback is bad", you could more reasonably say that in the past many poor designs managed to artificially inflate their specifications using excessive feedback, but the result was even worse sound than they started out with - and so, in those particular cases, feedback made tings worse.
Circuit characteristics like slew rate and open loop gain determine what level of feedback is appropriate or reasonable for a given circuit design, and must be taken into consideration; but, if they are properly considered, then feedback can improve the specifications and the sound quality - and avoid negative side effects. Unfortunately, once of the major down-sides of poor distortion specs is..... high distortion. While it may be true that inappropriately applying too much feedback can result in poor sound quality, allowing too much distortion to remain is also a problem. (I'm sorry, but while 0.5% of "benign distortion" may sound better than 0.05% of "nasty high order distortion", 10% of ANY kind of distortion is offensive and sounds nasty - at least to me - and many "low feedback designs" deliver exactly that.)
You also need to differentiate between "no feedback" and CERTAIN KINDS OF FEEDBACK. For example, virtually all tube designs, and even many solid state ones, can only reasonably apply a limited amount of GLOBAL FEEDBACK (feedback around the entire circuit), but may still apply plenty of LOCAL feedback at each individual stage, or between certain stages. Before you make generalizations about feedback, you should consider the fact that the venerable cathode follower buffer, which is used in the output stage of the majority of high end tube preamps, actually has INFINITE feedback, and an unbypassed cathode resistor is (literally) a textbook example of "local degenerative feedback". (And I wonder how many of those designs claim "no feedback" or "low feedback" in the marketing literature).
In simplest terms, it's VERY difficult to design a circuit that has acceptably low levels of distortion with no feedback at all.
About Output Impedance
I'm not prepared to go into a full explanation of what that means here (try an electronics textbook, or Google), but I will tell you that the output impedance of the XSP-1 is right around 1 kOhm. This means that you will see a slight drop in output LEVEL if you connect it to a load that is below about 10 kOhms, and in fact the level will drop by about 60% into a 600 Ohm load, but the distortion, frequency response, and noise level will remain the same - you'll just have to turn the volume up a bit higher to get the same output voltage level. In fact, the XSP-1 will probably drive pretty much any load with no problems - although the level will continue to drop (if you know how output impedance actually works, you can calculate the effect for any load). Any suggestion that this is a shortcoming or limitation is just silly; we could drop the output impedance to a few Ohms by cutting out a few fifty cent parts, but it would cause certain other things to not work as well.
Some devices actually have significant limitations on the amount of output current they can deliver. This, in turn, means that they will start distorting if you attach a load that has too low an impedance to them. (It would be proper to say that they have their specified output impedance only for loads above their stated minimum load; for loads below that, all bets are off, and they distort or fault.) The 600 Ohm minimum load specified for the XSP-1 is conservative; the XSP-1 can happily drive loads way below that, with low noise and distortion, but the output level will become too low to be useful.
Now, all of the discussion so far has assumed that the load we're driving is a pure resistor - and this is true for the input of pretty much EVERY modern amplifier or other analog audio device (but it isn't true for many really vintage amps and tube units - which may use input transformers). The output impedance of a device will interact with the impedance of the load device AND THE INTERCONNECT CABLE (which includes resistance, inductance, and capacitance) if the values are poorly chosen. Using the example of the XSP-1, if you were to connect your XSP-1 to your power amp using 1000 feet of shielded coax cable, the 1 kOhm output impedance of the XSP-1 would interact with the capacitance of the cable, and produce a (possibly noticeable) roll off of high frequencies (however this would be an unusual situation, and one that is not recommended). Many vintage tube preamps had output impedances of over 100k Ohms, and so did have problems with even normal length cables... but the XSP-1's 1 kOhm output impedance will have negligible interaction with any cable you're likely to encounter (unless it is very poorly designed or deliberately adds excess capacitance in an attempt to "tune" your audio signals).
About INPUT Impedance
When you talk about connecting other source devices (like DACs or Oppos) TO the XSP-1, you're talking about INPUT impedance. The output impedance of your DAC or Oppo interacts with the INPUT impedance of the XSP-1. Since the input impedance of the line level inputs on the XSP-1 is a very benign 47k Ohms, and is almost purely resistive, I can't imagine it's producing a significant interaction with any modern device like an Oppo or a DAC (unless the DAC in question had very unusual output circuitry, which a few oddball ones do). I've never played with one, but the Mytek DAC is a relatively "mainstream" product, so I'm sure it's pretty normal in this regard. I'm more inclined to put obvious differences down to the difference in room or speakers than to any sort of input interaction with the XSP-1.
dogsled: I have some questions regarding your intepretation. You see a low distortion spec and are concerned? It's hard to say if it is acheieved by a loop feedback or just kickbutt electronics would be my counter.
I am interested in what you said about the impedance: the XSP-1 gen 2 has slightly lower output impedance than the gen 2. But your DAC is already pretty high end. We experienced different results when trying different DAC's. For example for some wierd reason the dc-1 sounded recessed and somewhat bloated on the gen 2 unit (granted - I tried this in a different room than mine) so I do wonder if something to do with impedance is coming into play. The XSP-1 reccomends a minimum input impedance of 600 ohms which I suspect your mytek is outputting a lot lower than that.
The oppo 105 on the other hand ran very nicely with the XSP-1 and so for some reason did the XDA-2.
About Feedback.
It is true that usually, when you see very low distortion specs, they are achieved by using a lot of feedback - the reason being that feedback is a very effective way of reducing distortion. However, most of the misconceptions about how feedback relates to sound quality are based on largely outdated assumptions. There is nothing INHERENTLY bad about applying feedback. Feedback is only detrimental to sound quality when it is used IMPROPERLY. Unfortunately, it's easy to use feedback to make a poorly designed circuit "measure" better, which is why historically many circuits that had excellent specifications, but still managed to sound bad, turned out to have been designed with very high levels of feedback. (It was simply a combination of not understanding how other circuit characteristics affect the result of applying feedback, and of being unwilling or unable to avoid the temptation of "fixing the numbers" with insufficient regard for sound quality. Many of the problems experienced by early designs, like transient distortion, are relatively easy to detect and avoid - once you know what to look for. The original Dynaco PAT4 preamp was the poster child for a solid state preamp that "measured well but sounded awful".) Rather than claim "feedback is bad", you could more reasonably say that in the past many poor designs managed to artificially inflate their specifications using excessive feedback, but the result was even worse sound than they started out with - and so, in those particular cases, feedback made tings worse.
Circuit characteristics like slew rate and open loop gain determine what level of feedback is appropriate or reasonable for a given circuit design, and must be taken into consideration; but, if they are properly considered, then feedback can improve the specifications and the sound quality - and avoid negative side effects. Unfortunately, once of the major down-sides of poor distortion specs is..... high distortion. While it may be true that inappropriately applying too much feedback can result in poor sound quality, allowing too much distortion to remain is also a problem. (I'm sorry, but while 0.5% of "benign distortion" may sound better than 0.05% of "nasty high order distortion", 10% of ANY kind of distortion is offensive and sounds nasty - at least to me - and many "low feedback designs" deliver exactly that.)
You also need to differentiate between "no feedback" and CERTAIN KINDS OF FEEDBACK. For example, virtually all tube designs, and even many solid state ones, can only reasonably apply a limited amount of GLOBAL FEEDBACK (feedback around the entire circuit), but may still apply plenty of LOCAL feedback at each individual stage, or between certain stages. Before you make generalizations about feedback, you should consider the fact that the venerable cathode follower buffer, which is used in the output stage of the majority of high end tube preamps, actually has INFINITE feedback, and an unbypassed cathode resistor is (literally) a textbook example of "local degenerative feedback". (And I wonder how many of those designs claim "no feedback" or "low feedback" in the marketing literature).
In simplest terms, it's VERY difficult to design a circuit that has acceptably low levels of distortion with no feedback at all.
About Output Impedance
I'm not prepared to go into a full explanation of what that means here (try an electronics textbook, or Google), but I will tell you that the output impedance of the XSP-1 is right around 1 kOhm. This means that you will see a slight drop in output LEVEL if you connect it to a load that is below about 10 kOhms, and in fact the level will drop by about 60% into a 600 Ohm load, but the distortion, frequency response, and noise level will remain the same - you'll just have to turn the volume up a bit higher to get the same output voltage level. In fact, the XSP-1 will probably drive pretty much any load with no problems - although the level will continue to drop (if you know how output impedance actually works, you can calculate the effect for any load). Any suggestion that this is a shortcoming or limitation is just silly; we could drop the output impedance to a few Ohms by cutting out a few fifty cent parts, but it would cause certain other things to not work as well.
Some devices actually have significant limitations on the amount of output current they can deliver. This, in turn, means that they will start distorting if you attach a load that has too low an impedance to them. (It would be proper to say that they have their specified output impedance only for loads above their stated minimum load; for loads below that, all bets are off, and they distort or fault.) The 600 Ohm minimum load specified for the XSP-1 is conservative; the XSP-1 can happily drive loads way below that, with low noise and distortion, but the output level will become too low to be useful.
Now, all of the discussion so far has assumed that the load we're driving is a pure resistor - and this is true for the input of pretty much EVERY modern amplifier or other analog audio device (but it isn't true for many really vintage amps and tube units - which may use input transformers). The output impedance of a device will interact with the impedance of the load device AND THE INTERCONNECT CABLE (which includes resistance, inductance, and capacitance) if the values are poorly chosen. Using the example of the XSP-1, if you were to connect your XSP-1 to your power amp using 1000 feet of shielded coax cable, the 1 kOhm output impedance of the XSP-1 would interact with the capacitance of the cable, and produce a (possibly noticeable) roll off of high frequencies (however this would be an unusual situation, and one that is not recommended). Many vintage tube preamps had output impedances of over 100k Ohms, and so did have problems with even normal length cables... but the XSP-1's 1 kOhm output impedance will have negligible interaction with any cable you're likely to encounter (unless it is very poorly designed or deliberately adds excess capacitance in an attempt to "tune" your audio signals).
About INPUT Impedance
When you talk about connecting other source devices (like DACs or Oppos) TO the XSP-1, you're talking about INPUT impedance. The output impedance of your DAC or Oppo interacts with the INPUT impedance of the XSP-1. Since the input impedance of the line level inputs on the XSP-1 is a very benign 47k Ohms, and is almost purely resistive, I can't imagine it's producing a significant interaction with any modern device like an Oppo or a DAC (unless the DAC in question had very unusual output circuitry, which a few oddball ones do). I've never played with one, but the Mytek DAC is a relatively "mainstream" product, so I'm sure it's pretty normal in this regard. I'm more inclined to put obvious differences down to the difference in room or speakers than to any sort of input interaction with the XSP-1.
I am interested in what you said about the impedance: the XSP-1 gen 2 has slightly lower output impedance than the gen 2. But your DAC is already pretty high end. We experienced different results when trying different DAC's. For example for some wierd reason the dc-1 sounded recessed and somewhat bloated on the gen 2 unit (granted - I tried this in a different room than mine) so I do wonder if something to do with impedance is coming into play. The XSP-1 reccomends a minimum input impedance of 600 ohms which I suspect your mytek is outputting a lot lower than that.
The oppo 105 on the other hand ran very nicely with the XSP-1 and so for some reason did the XDA-2.
