Marantz PM-11S3 Integrated Amplifier Review

[Irvrobinson]

I'd be interested in knowing how balanced topology reduces distortion.

Care to share your wisdom on this?

Are you saying balanced topologies don't reduce even-order harmonics? That balanced topologies only reduce noise?

 

[avengineer]

Are you saying balanced topologies don't reduce even-order harmonics? That balanced topologies only reduce noise?

The original statement from Gene said, "With all things being equal, I would take a fully complimentary design from input to output ANY day over its single ended equivalent. It's not just for noise immunity but for the distortion reducing benefits of the topology."

All I said was, "I'd be interested in knowing how balanced topology reduces distortion."

I did not specifically say that balanced topologies reduce noise. My original quote is, "It should be noted that the entire point of a balanced interconnection is common-mode noise rejection."

There is a difference between a topology reducing noise and rejecting noise. In fact, it's likely that a completely balanced system, source to input, will have a slightly higher noise floor than single ended because of all the additional electronics required, though the final S/N ratio may be better.

There is also a significant difference between "balanced" and "differential" and "fully complimentary", which many people seem to lump together as the same concept. I suspect that's where this discussion is being de-railed.

However, in the context of this review, I was making my comment regarding the provision of a "balanced" input connector, and the lack of a common-mode rejection measurement. I was not referencing internal circuit topology, only the input itself. I still stand by statement about the purpose of a "balanced" connection. And, by extension, if the purpose is noise rejection (or noise pickup immunity), then that parameter should be measured. Many so-called "balanced inputs" are nothing of the kind and provide only marginal CMRR, which is why that key parameter is almost never specified, and if it is, it's stated as a single number, when, in fact, CMRR usually varies significantly with frequency. The point is, if a feature is provided on a high-end component, and that feature is somehow valued by the consumer or touted as an advantage by the manufacturer, then it's primary function should be tested. Would you pay extra for a turbocharger on a car without at least some idea of how much additional horsepower it could provide?

I'm still interested in learning how balanced topology reduces distortion over a single-ended equivalent.

And just to add this, my comments are intended as constructive criticism. I'm sorry if they are not looked upon that way. I'd like to see the goal of the pursuit of "Truth" continue, and that's why I offer these comments. My comments are not meant to be argumentative, or to denigrate anyone, though I'm sorry to see that someone felt the need to hit my reputation because I offer criticism. I do not intend to be "argumentative".

 

[gene]

The original statement from Gene said, "With all things being equal, I would take a fully complimentary design from input to output ANY day over its single ended equivalent. It's not just for noise immunity but for the distortion reducing benefits of the topology."

All I said was, "I'd be interested in knowing how balanced topology reduces distortion."

I did not specifically say that balanced topologies reduce noise. My original quote is, "It should be noted that the entire point of a balanced interconnection is common-mode noise rejection."

There is a difference between a topology reducing noise and rejecting noise. In fact, it's likely that a completely balanced system, source to input, will have a slightly higher noise floor than single ended because of all the additional electronics required, though the final S/N ratio may be better.

There is also a significant difference between "balanced" and "differential" and "fully complimentary", which many people seem to lump together as the same concept. I suspect that's where this discussion is being de-railed.

However, in the context of this review, I was making my comment regarding the provision of a "balanced" input connector, and the lack of a common-mode rejection measurement. I was not referencing internal circuit topology, only the input itself. I still stand by statement about the purpose of a "balanced" connection. And, by extension, if the purpose is noise rejection (or noise pickup immunity), then that parameter should be measured. Many so-called "balanced inputs" are nothing of the kind and provide only marginal CMRR, which is why that key parameter is almost never specified, and if it is, it's stated as a single number, when, in fact, CMRR usually varies significantly with frequency. The point is, if a feature is provided on a high-end component, and that feature is somehow valued by the consumer or touted as an advantage by the manufacturer, then it's primary function should be tested. Would you pay extra for a turbocharger on a car without at least some idea of how much additional horsepower it could provide?

I'm still interested in learning how balanced topology reduces distortion over a single-ended equivalent.

And just to add this, my comments are intended as constructive criticism. I'm sorry if they are not looked upon that way. I'd like to see the goal of the pursuit of "Truth" continue, and that's why I offer these comments. My comments are not meant to be argumentative, or to denigrate anyone, though I'm sorry to see that someone felt the need to hit my reputation because I offer criticism. I do not intend to be "argumentative".

If you truly are an EE then you wouldn't be asking how a differential topology reduces distortion over an equivalent Single Ended design. There are plenty of EE telecom books that go into great detail about this but even a junior level Curcuits I book from college broaches the subject. When I get back to the office I can even give a few references for recommended reading if you like.

 

[gene]

Thanks, Gene. I'm only giving constructive criticism. If it's not welcome, then perhaps it at least informs other readers in some way.

This would be the second time you've suggested I start my own AV publication, though. Do you really want that competition?

Soery but your posts come off very abrasively. As for competition, good luck with that. This is a small and tough field to thrive in. The print mags backed by million dollar corporations have a sophomoric understanding of SEO which thankfully keeps us at the top of the food chain in traffic and reach online.

 

[avengineer]

If you truly are an EE then you wouldn't be asking how a differential topology reduces distortion over an equivalent Single Ended design. There are plenty of EE telecom books that go into great detail about this but even a junior level Curcuits I book from college broaches the subject. When I get back to the office I can even give a few references for recommended reading if you like.

Gene,

I don't believe I've ever claimed to hold a EE degree. But that degree is not always an indicator of knowledge, is it? I once stood next to a EE while he put a scope probe on the inverting input of an opamp and thought it was broken because he saw almost no signal there. I do know what goes into that degree, and no it's not a proof of all knowledge, especially about high quality audio. There are many EEs, of course, no doubt yourself included, that do possess a deep knowledge of audio circuits. It's not an indicator either way. There are also many pre-eminant figures in the audio industry that don't hold EE degrees.

Again, we are probably confusing several concepts. I understand the differences of internal circuit design topologies. The point here is the advantages of a balanced interface of one device to another, not internal topology. There is a difference, as you know. I'm fairly sure that when balanced technology is referenced in a telecom book, it is in the context of interface. In that context, I would like to see how a balanced interface reduces distortion, so long as we define "distortion" in the traditional way, and don't include the contemporary concept of bundling noise and frequency response into the term. Fair enough?

I'll happily look up any EE Telecom, junior-level Circuits, or even grade-school electricity reference you cite, and learn from it. I love to learn.

 

[avengineer]

Soery but your posts come off very abrasively. As for competition, good luck with that. This is a small and tough field to thrive in. The print mags backed by million dollar corporations have a sophomoric understanding of SEO which thankfully keeps us at the top of the food chain in traffic and reach online.

You did see the smiley-face, right? Seriously, I'm not going to be your competition.

 

[Irvrobinson]

The original statement from Gene said, "With all things being equal, I would take a fully complimentary design from input to output ANY day over its single ended equivalent. It's not just for noise immunity but for the distortion reducing benefits of the topology."

All I said was, "I'd be interested in knowing how balanced topology reduces distortion."

It is a well-known concept that even-order harmonics are often reduced by several decibels in balanced audio circuits, and I suspect you know this, and that you have an agenda you are trying to subtly reveal. Is it the mathematics of the cancellation you want to see? Frankly, I'm not sure what your point is.

Just for the record, I agree with Gene, I'd rather have well-designed fully balanced designs in amplifiers over non-balanced designs.

I did not specifically say that balanced topologies reduce noise. My original quote is, "It should be noted that the entire point of a balanced interconnection is common-mode noise rejection."

Yes, I know, but I interpreted your question, perhaps inappropriately, that you did not believe balanced topologies could reduce distortion. Gene and I had a discussion about balanced topologies in another thread, and in his defense of his firm position he produced a compelling plot of the distortion of an Emotiva amp in bridged and unbridged modes, with the bridged mode demonstrating significantly lower distortion. It was very compelling evidence, and I found that his premise was also supported by measurements of the Pass Labs amp by Stereophile, which have a very unusual "balanced single-ended" topology, and the distortion of the amp was to be mostly in the 3rd harmonic, which would be exactly what you would expect if second-order harmonic distortion cancellation was occurring. (What Stereophile observed implies a rather extraordinary level of matching in the amp components, which makes Pass Labs amps even more intriguing.)

There is a difference between a topology reducing noise and rejecting noise. In fact, it's likely that a completely balanced system, source to input, will have a slightly higher noise floor than single ended because of all the additional electronics required, though the final S/N ratio may be better.

In the context of normal casual conversation you're just being difficult, IMO. Many engineers discuss the noise rejection of balanced circuits as "reducing noise", when what they really mean is that the cancellation of common mode noise results in lower noise at the outputs. As for your statement that balanced designs "will have" a higher noise floor than a single-ended design, I'm not seeing it. What additionally circuitry? Twice the power supply capacity? The rest of the circuitry is mostly just a replication of what you need for a single-ended design, and with proper matching you should get quite a bit of noise and even-order harmonic distortion cancellation.

There is also a significant difference between "balanced" and "differential" and "fully complimentary", which many people seem to lump together as the same concept. I suspect that's where this discussion is being de-railed.

Doubtful.

However, in the context of this review, I was making my comment regarding the provision of a "balanced" input connector, and the lack of a common-mode rejection measurement. I was not referencing internal circuit topology, only the input itself. I still stand by statement about the purpose of a "balanced" connection. And, by extension, if the purpose is noise rejection (or noise pickup immunity), then that parameter should be measured. Many so-called "balanced inputs" are nothing of the kind and provide only marginal CMRR, which is why that key parameter is almost never specified, and if it is, it's stated as a single number, when, in fact, CMRR usually varies significantly with frequency. The point is, if a feature is provided on a high-end component, and that feature is somehow valued by the consumer or touted as an advantage by the manufacturer, then it's primary function should be tested. Would you pay extra for a turbocharger on a car without at least some idea of how much additional horsepower it could provide?

I agree that some components provide XLR inputs and outputs that are supported with little more than phase splitter circuitry, but in looking over your original comments this isn't the point that you appeared to be making. Your point appeared to be that balanced circuitry had a lot more components and were often poorly done, so therefore balanced topologies were of marginal value. I disagree, and I'm pretty sure that's what Gene responded to.

I'm still interested in learning how balanced topology reduces distortion over a single-ended equivalent.

By the cancellation of even-order harmonics.

 

[avengineer]

It is a well-known concept that even-order harmonics are often reduced by several decibels in balanced audio circuits, and I suspect you know this, and that you have an agenda you are trying to subtly reveal. Is it the mathematics of the cancellation you want to see? Frankly, I'm not sure what your point is.

Ok, it's now very clear that I haven't been very clear. I'm NOT talking about internal circuit topology. I'm talking about the balanced interface itself. I do understand the math.

Yes, I know, but I interpreted your question, perhaps inappropriately, that you did not believe balanced topologies could reduce distortion. Gene and I had a discussion about balanced topologies in another thread, and in his defense of his firm position he produced a compelling plot of the distortion of an Emotiva amp in bridged and unbridged modes, with the bridged mode demonstrating significantly lower distortion. It was very compelling evidence, and I found that his premise was also supported by measurements of the Pass Labs amp by Stereophile, which have a very unusual "balanced single-ended" topology, and the distortion of the amp was to be mostly in the 3rd harmonic, which would be exactly what you would expect if second-order harmonic distortion cancellation was occurring. (What Stereophile observed implies a rather extraordinary level of matching in the amp components, which makes Pass Labs amps even more intriguing.)

Again, the fact that INTERNALLY, a so-called balanced topology has the potential to reduced certain types of distortion is correct. No argument. And, that's NOT what I'm referring to.

In the context of normal casual conversation you're just being difficult, IMO. Many engineers discuss the noise rejection of balanced circuits as "reducing noise", when what they really mean is that the cancellation of common mode noise results in lower noise at the outputs. As for your statement that balanced designs "will have" a higher noise floor than a single-ended design, I'm not seeing it. What additionally circuitry? Twice the power supply capacity? The rest of the circuitry is mostly just a replication of what you need for a single-ended design, and with proper matching you should get quite a bit of noise and even-order harmonic distortion cancellation.

Again, I'm talking interface, you guys are hung up on internal topology. But, so long as we're in there... balanced topology can only cancel noise and distortion in the common mode. In internal balanced topology you have many different sources of uncorrelated noise in both paths, and so they cannot cancel, in fact they add on the order of 3dB. Distortion products caused by similar mechanisms that are similar in both halves can, if exactly identical, be canceled, provided they occur in the common mode, otherwise they are differential and out they come.

In a classic instrumentation input stage there are at least 3 times as many components as a classic single ended input because you need a voltage follower on each of the inverting and non-inverting inputs, then a diff amp. For a classic single ended input, you'd only need a voltage follower, or perhaps a single stage with a bit of gain. The additional stages of the instrumentation amp have the potential to increase noise because the noise sources are uncorrelated and thus appear as differential signals. If we consider a true balanced input stage that involves a transformer, then we have the possibility of a better source impedance match to the following gain stage, which could actually reduce the total system noise, but we then have whatever distortion products are created by the transformer, which may be minimal, but are always measurable, plus bandwidth and phase penalties.

For a balanced output stage, the two common topologies are a phase splitter with output buffers or a unity-gain inverting amplifier doing the same job, or a transformer. In the case case of the phase splitter or inverting amp, we've doubled the component count and doubled the noise sources, all of which now create uncorrelated noise which will show up as a differential signal, not a common-mode signal that could be cancelled at the receiving input. The second common balancing method is to use a transformer, which has again the potential for better impedance match and possibly lower noise, but at the expense of distortion, phase and bandwidth penalties.

By the way, in the process of any of these balancing input our output stages, distortion products that are produced are mostly differential, with very little happening in common-mode.

Internally in an amplifier, the story is a bit different, and there are common-mode distortion mechanisms, and there can be a distortion advantage.

One more time, I'm not talking about an internal amplifier design, in fact, in the context of this review, the device is single ended with a balanced interface connection. That is the specific case we are talking about here, you both have extended the discussion to internal circuit topology that is irrelevant in the review of this device.

I agree that some components provide XLR inputs and outputs that are supported with little more than phase splitter circuitry, but in looking over your original comments this isn't the point that you appeared to be making. Your point appeared to be that balanced circuitry had a lot more components and were often poorly done, so therefore balanced topologies were of marginal value. I disagree, and I'm pretty sure that's what Gene responded to.

Again, I apologize for being so vague. My only point in mentioning the balanced interface was to question its performance in its primary function on this single-ended device: common-mode noise rejection.

By the cancellation of even-order harmonics.

And yet again...we are not in disagreement about internal amplifier design and potential benefits. I'm only talking about the interface.

Since I've already irritated the good ol' boys club here, I might as well close by dropping this bomb: I'm fairly sure the distortion products that are reduced in a fully balanced amplifier design are even-order, probably predominantly second-order. In the world of audible harmonic distortion, these would be the least audible, with case studies showing second-order THD being difficult to reliably detect in a DBT at levels as high as 3%. So, while reducing even-order harmonic distortion by virtue of a balanced amp topology may result in a measurable improvement, it won't be an audible improvement. I call into question the logic of spending additional money on an inaudible improvement when spending those dollars on better transducers might result in a better ROI.

Go ahead, guys, blast away, discredit me, hit my rep, whatever. I'm pretty much done here anyway. I was trying to advance the understanding of Truth in audio (that was, in fact, the "hidden agenda"...sheesh), but I'm seen as argumentative. Sorry, guys...not my intention. I just don't tend to sugar-coat technical discussions. They are non-emotional topics for me, sorry if they are not so for you.

I guess this whole thing was a bad idea.

 

[RichB]

That is not the first time that I have read that a fully balanced amp reduces even order harmonics and not the more objectionable odd order.

i believe that balanced amps may be more efficient.

- Rich

 

[gene]

Since I've already irritated the good ol' boys club here, I might as well close by dropping this bomb: I'm fairly sure the distortion products that are reduced in a fully balanced amplifier design are even-order, probably predominantly second-order. In the world of audible harmonic distortion, these would be the least audible, with case studies showing second-order THD being difficult to reliably detect in a DBT at levels as high as 3%. So, while reducing even-order harmonic distortion by virtue of a balanced amp topology may result in a measurable improvement, it won't be an audible improvement. I call into question the logic of spending additional money on an inaudible improvement when spending those dollars on better transducers might result in a better ROI.

I honestly don't have the time to debate every point in your posts. You do raise some valid criticisms but your delivery can use some improvement.

Be careful about over objectifying audio. Usually people fall in the trap of using too simplistic of testing procedure and too small of a sample size to draw conclusions. Most amplifier distortion measurements are pretty useless at correlating audibility. Most magazines simply do a 1kHz instantaneous sweep test and miss a whole slew of problems in the process. Think back a few years ago when the print magazines were all going gaga over ICE amps in Pioneer receivers. It wasn't until I and a select few other industry professionals (NOT reviewers) started measuring them and seeing the severe limitations they had into 4 ohm loads above 2kHz that people started to see the light. I had loudspeaker companies call me and thank me for bringing this topic out in the open since it was affecting returns of their 4 ohm speakers. Pioneer engineers thanked me in the elevator at CEDIA as a result of my testing and have since then moved on to different amp modules.

Once you start measuring full bandwidth multi-tone distortion instead just a single tone at a single frequency, you start seeing more interesting things going on. I have also been able to hear clear sonic differences between different amps both not being driven into distortion and both level matched if the load impedance was complex enough and the speakers revealing enough.

I stopped oversimplifying conclusions based on limited testing and now use measurements as a starting point to snuff out design issues rather than to make "objective" conclusions that one product sounds better than another b/c of measurements.

This is where the NRC measurements really show short comings, especially in distortion measurements, and as a result people draw false conclusions about product performance and some designers make poor decisions like running a 5" midrange fullrange with no crossover component b/c an LMS sweep didn't reveal any distortion. That is another topic for another thread and perhaps a formal article however.

I wish I had more time these days to dedicate to article topics like this (amplifier distortion, balanced vs single ended, etc). I may collaborate with my friends from Audio Precision however as this topic seems to come up often on our forum.

Two books I recommend reading are:

• Noise Reduction Techniques in Electronic Systems by Henry W. Ott
• The Art of Electronics by Paul Horowitz and Winfield Hill

 

[gene]

That is not the first time that I have read that a fully balanced amp reduces even order harmonics and not the more objectionable odd order.

i believe that balanced amps may be more efficient.

- Rich

They kinda are b/c of how the power supply is done. I discuss this in my Emotiva review if you would like to read the design section:
Emotiva XPR-1 Mono Amplifier Review | Audioholics

 

[Irvrobinson]

In internal balanced topology you have many different sources of uncorrelated noise in both paths, and so they cannot cancel, in fact they add on the order of 3dB.

Wait, what? You're all confused. The uncorrelated noise may indeed double, because there are two paths, but that doesn't mean that the overall SNR is degrading by 3db, it just means that that minute component of the overall noise level may double.

 

[avengineer]

Thanks for the suggestions, Gene.

I haven't characterized distortion performance of a device or system with a single tone at a single frequency once in my life, not since my first THD measurement in 1974. Even doing the old broadcast "audio proof" required many tones and many levels. In fact, after loosing interest in traditional THD measurements in the late '70s, I began using a spectrum analyzer to discern actual frequency distribution of distortion products, which led to CCIF-style IMD, and all down that convoluted TIM road of the 1980s, all pretty much pre FFT, at least as far as common usage goes, and that was also pre-Audio Precision when they all worked for Tek.

Thanks for the book suggestions. I own both already. Ott, chapter 4 is exactly what I've been trying to talk about re: balanced interface. Ott, ch 9, "Active Device Noise" underscores why a balanced topology will be noisier than the single-ended equivalent.

"The Art of Electronics" is pretty basic stuff, not sure why you thought that was important here. I had to dig out my copy to refresh my memory of what it was about, but now I see why it wasn't where I would look at it much. Was there a particular chapter?

Its refreshing to see your interest in detailed testing. I agree that oversimplified conclusions based on limited testing is misleading, and in fact, my "suggestions" re: additional tests or proper phono preamp input termination are an effort to move you beyond the cursory for those specific tests. In fact, I would encourage full-bandwidth multi-tone distortion testing (IMD) on all audio product reviews, a well as plots of THD+N vs frequency vs level as a 3d plot, just for starters.

If a listening test seems to reveal a particular quality, I would also encourage doing a DBT against another similar device to determine if that observation is real or the result of expectation bias. After decades of ABX/DBT testing I'm satisfied that we don't always hear what we think we hear, and that perception is incredibly easy to influence.

Yeah, I know, "go start your own magazine". Not going to happen, but if any of this helps improve this one, that's fine. Or, I'll just accept the label of "crotchety old man".

 

[avengineer]

Wait, what? You're all confused. The uncorrelated noise may indeed double, because there are two paths, but that doesn't mean that the overall SNR is degrading by 3db, it just means that that minute component of the overall noise level may double.

No, I'm not confused at all. If you could (an you really can't) compare the single-ended version of an amplifier with a fully balanced, gain compensated version, there are more than twice the uncorrelated noise sources. The problem is, you can't make that comparison because you can't just switch operating modes of something like a commercially made amp and expect to have sliced off all the circuitry involved in fully balancing it. That would be a huge assumption that would require close examination of the schematic, which I doubt anyone publishes these days. But, if you could prove that's what's going on, and make some valid measurements with proper input termination (no dead shorts, no low Z test gear), and prove that the single ended version is noisier, I'll stand corrected. Don't forget to watch for a 6dB gain change, though, and comp it out when you do the S/N test.

I'm not sure what other noise sources you're thinking would dominate in an amplifier.

Now, if we're talking total system noise...source to final output...that's a different question with a different answer. The confusion, I think, remains that I've tried to separate the interface portion of the question from total topology. My purpose in doing so was driven by the existence of a balanced input connector on an otherwise single-ended amplifier...the Marantz in this review.

It wen't all wonky from that point on.

 

[Irvrobinson]

No, I'm not confused at all. If you could (an you really can't) compare the single-ended version of an amplifier with a fully balanced, gain compensated version, there are more than twice the uncorrelated noise sources. The problem is, you can't make that comparison because you can't just switch operating modes of something like a commercially made amp and expect to have sliced off all the circuitry involved in fully balancing it. That would be a huge assumption that would require close examination of the schematic, which I doubt anyone publishes these days. But, if you could prove that's what's going on, and make some valid measurements with proper input termination (no dead shorts, no low Z test gear), and prove that the single ended version is noisier, I'll stand corrected. Don't forget to watch for a 6dB gain change, though, and comp it out when you do the S/N test.

Good component matching is going to mean that the circuits in both phases are very close to identical, so that the CMRR is high. I must be misinterpreting you, because you seem to be arguing that single-ended designs will have inherently lower noise than balanced designs.

I'm not sure what other noise sources you're thinking would dominate in an amplifier.

The noise inherent in the amplification signal path itself.

Now, if we're talking total system noise...source to final output...that's a different question with a different answer. The confusion, I think, remains that I've tried to separate the interface portion of the question from total topology. My purpose in doing so was driven by the existence of a balanced input connector on an otherwise single-ended amplifier...the Marantz in this review.

It wen't all wonky from that point on.

This was your question in post #20 that started the balanced circuitry discussion:

I'd be interested in knowing how balanced topology reduces distortion.

Perhaps we misinterpreted you, but topology is a circuitry term, not typically used to describe connectors.

 

[avengineer]

Good component matching is going to mean that the circuits in both phases are very close to identical, so that the CMRR is high. I must be misinterpreting you, because you seem to be arguing that single-ended designs will have inherently lower noise than balanced designs.

Nope, you've got that one right. Even with precision beta match in transistors, you can't match noise, only noise level. Two transistors (matched) are two uncorrelated noise sources which will add as differential signals.

The noise inherent in the amplification signal path itself.

Yup...that's exactly what I'm saying.

This was your question in post #20 that started the balanced circuitry discussion:

Perhaps we misinterpreted you, but topology is a circuitry term, not typically used to describe connectors.

[/quote]

Nope, you've got it. I'm referring to balanced "topology" when referencing the internal circuitry of an amplifier, and the wild claims that a balanced "topology" (internally) would result in lower noise.

As to connectors, you're not reading my posts very well... A "balanced" style connector on a component implies a balanced input interface, and the ONLY POINT of a balanced INPUT is common-mode noise rejection, a parameter that is typically not specified, or under-specified, and should be tested in a review so consumers can have an indication of the performance of that input's sole function in life. A balanced connector makes NO implication about the INTERNAL TOPOLOGY of the device, most of which, in point of fact, are single ended internally, as it the Marantz in question. IF the device were fully balanced, input interface to output interface, then and only then would the internal topology come into play, albeit minimally.

Then...Gene claimed he'd generally prefer balanced or differential over single-ended because of the distortion cancellation. Sorry, Gene, if I've misquoted. Outside of a small degree of predominantly second order (crossover) distortion cancellation - because it occurs in common-mode (a type of distortion that is least audible), I've yet to see any evidence of any significant distortion cancellation by the INTERNAL TOPOLOGY of an amplifier or device, and certainly none by virtue of a balanced interface to the outside world. Distortion caused by nonlinearities will not occur in the common-mode and cannot be cancelled.

 

[gene]

Then...Gene claimed he'd generally prefer balanced or differential over single-ended because of the distortion cancellation. Sorry, Gene, if I've misquoted. Outside of a small degree of predominantly second order (crossover) distortion cancellation - because it occurs in common-mode (a type of distortion that is least audible), I've yet to see any evidence of any significant distortion cancellation by the INTERNAL TOPOLOGY of an amplifier or device, and certainly none by virtue of a balanced interface to the outside world. Distortion caused by nonlinearities will not occur in the common-mode and cannot be cancelled.

You obviously have never worked in the Telecom industry. Every competent design in the Telcom world is fully differential for two reasons:

• lower noise
• lower distortion

Sorry I don't have time to sit here and debate you on it but these are engineering facts. DSL over POTS would have NEVER worked with single ended, unbalanced designs. Every competent engineer knows this.

You have a talent for cherry picking data or quotes to push some weird agenda against it. I don't understand why, and quite frankly I really don't care.

 

[gene]

Thanks for the book suggestions. I own both already. Ott, chapter 4 is exactly what I've been trying to talk about re: balanced interface. Ott, ch 9, "Active Device Noise" underscores why a balanced topology will be noisier than the single-ended equivalent.

I happen to be friends with Henry Ott and I think he would find it amusing how you are misquoting his book. In fact, he's even written and peer reviewed many cable articles on this site. He and I had many discussions on the merits of fully differential designs in audio.

Let me just quote a small excerpt from his book on page 116: "Balancing is an often overlooked-though in many cases cost-effective-noise reduction technique, which may be used in conjunction with shielding when noise must be reduced below the level obtainable with shielding alone."

And check out Henry's little App note inspired by one of the articles we coauthored on this site many years ago:
http://www.hottconsultants.com/pdf_files/Audio Interconnections.pdf

Therefore we can conclude that for the situation described above, regardless of the signal level the balanced interface will have 60 dB or more, depending upon the degree of balance achieved, less noise than the unbalanced interface

I honestly cannot believe we are still having this discussion, especially given your claimed background.

 

[Irvrobinson]

Nope, you've got it.

Well, okay, the show's over. I'll just agree to disagree and be quiet. I'm not sure that balanced circuitry is always cost effective or justified for audio equipment, especially for unity-gain, line-level, IC-based components, but I'm on the same page Gene is that balanced designs are always better than single-ended versions, when properly implemented. Let's leave it at that.

 

[avengineer]

You obviously have never worked in the Telecom industry. Every competent design in the Telcom world is fully differential for two reasons:

• lower noise
• lower distortion

I'm not sure why you say things like this. If it's an attempt to discredit me, so be it. Or, it might be an attempt at baiting me to revealing more about my careers or education, which you would then diminish and again use to discredit me. Not sure why you'd mention working in the Telecom industry (talk about "cherry picking"!), but since you did, I'll play another round with you.

First things first. Your comment "obviously" is a retort in apparent opposition to my statements. You clearly (or should I say, "Obviously"?) don't understand my statements at all.

I agree that balanced interfaces, especially in the Telecom industry, played a vital role in canceling common-mode noise on telephone lines. Nothing I've posted in this thread so far would contradict that...so long as you read my posts carefully. All I've ever said was, common-mode noise rejection/cancellation is the entire point. However the above second bullet point, "lower distortion" is completely unfounded in the discussion of a balanced or differential interface between devices. And, so far, you haven't presented any sound arguments or proofs to the contrary. In fact, you oddly blend the discussion of balanced interface with balanced amplifier topology. Gene, please understand, there is a difference, you can have one without the other. And I fully acknowledge that a balanced amplifier topology can cancel some distortion products, but I also acknowledge these to be predominantly second-order, and then other even-order products that are well known to be the least audible of all types of harmonic distortion. I therefore do not regard the internal amplifier fully-balanced topology as highly as you, especially since it more than doubles the number of noise sources in the system. If you could compare amplifiers with identical circuitry, but one fully balanced and the other single ended, you would find the best you could ever hope to do is a 3dB theoretical improvement in internal amplifier noise, but in practice, likely the single-ended version would be the quieter one.

Now, over to Telecom. Today, telecom networks are all switched digital systems, or VOIP systems, neither of which actually transmit audio, so the use of balanced audio circuits in the typical phone network is now confined to the "last mile", the last remaining copper pairs to subscribers. That loop is indeed basically balanced, but the device at the opposite end of that loop from the subscriber ends all of that. At AT&T, the system I'm most familiar with, local copper pairs end up at a device a short distance away (AT&T's term is VRAD), where the copper hits a card with 64 possible line input pairs. That card provides signaling, ring voltage, etc., for the subscriber, and actually can handle legacy phones right down to rotary dial. There are many cards in a VRAD, but at the end of it all sits an optical fiber that runs between the VRAD and Central Office, where calls enter the AT&T "cloud" as data. See, we lost those balanced audio lines back at the input to the VRAD cards. Not to say they aren't important, they are, but that's where it ends. And, hate to break it to you, after the input stage those cards are fully single ended right up to the bit stream output.

To really see Telco use balanced audio lines and interfacing we have to hop into the way-back machine to at least the 1980s, where we find the vestiges of analog phone systems. Perhaps the most interesting application of balanced interface technology would be in the program equalizer applied to broadcast lines that could be not only equalized for flat response to beyond 15KHz, but purchased "balanced for stereo", with precision gain and phase match. Now, the input to that program equalizer was a transformer - a balanced interface. But, on the equalizer side, the circuit was unbalanced (and passive!) while it did it's EQ trick, then the output was presented to another transformer for a balanced output to the customer's equipment. Even non-equalized program loops used simple 111C "repeat coils" to insure a balanced termination at the customer's equipment, sending or receiving. Telco never (ok, hardly ever) used shielded wire, and as you well know, that shield is electrostatic only anyway and has nothing whatever to do with reducing inductive coupling from one pair to another. And, Telco also has this habit of bundling hundreds of pairs of wire into one huge cable, so the chances for inductive and capacitive coupling are huge. Hence the absolute requirement for balanced lines, line drivers, and line receivers. You had a 15KHz broadcast line with an 80dB dynamic range right next to a POTS line with 90VAC of ring voltage and switching transients on it. Yes, balanced interface saved the day.

But that's all gone now. No Telecom company offers analog program circuits any more, they're all digital. And just before that, they were "carrier" circuits. The circuits themselves carried no audio, are certainly balanced for crosstalk reduction.

So, what's the point? Telecom uses balanced lines, drivers, and receivers. But, Gene, once the signal enters terminal equipment, the signal is processed by single-ended electronics! Not only is there no need for a fully balanced active system at that point, there's no desire to pay for it.

Do you think we could agree on this topic...even a little?

Sorry I don't have time to sit here and debate you on it but these are engineering facts. DSL over POTS would have NEVER worked with single ended, unbalanced designs. Every competent engineer knows this.

Thanks, finally a compliment! I do understand that! Does that make me a competent engineer in your eyes?

You have a talent for cherry picking data or quotes to push some weird agenda against it. I don't understand why, and quite frankly I really don't care.

Hmmm...so what would you call the Telecom thing?

Oh, and by the way, still no evidence that balanced interface reduces distortion.

And...I'm on to your next message! Woopee.

edit: Just read your bio...now I understand where the Telecom stuff came from.