Yep. The Musical Fidelity A1, designed by Tim de Paravicini. My friend with the "golden ears" paid a lot of money to have his revised.
It sounds like at tube amp to my "wooden" ears
Jörg.
The Levinson ML2, Levinson 27.5, Pass Aleph 0 come to mind.
Why Tube Amps Sound Different (and better) Than SS Amps
- Thread starter mep
- Start date
- Status
The Levinson ML2, Levinson 27.5, Pass Aleph 0 come to mind.
It's hard to believe this thread is becoming active again. I started this thread after being away from tubes for over two years because I was 'mad' at them and needed a break. This happened around 2011 and I sold all of my tube gear except for my Ampex 350s and went SS (scared solid state?). I bought all Krell gear from yesteryear-KSA-250 amp, first a KBL linestage which was replaced by the KRC-HR, and the KPE REF phono stage. I can and did live happily with that setup for several years (I still have the KPE REF phono stage). And then a pair of tube amps landed back in my system when I did the review of the Raven Silhouette monoblocks. I was smitten once again by all of the things that tube amps do right. I'm also a big fan of the KT-120/KT-150 tubes based on my limited experience with both of them.
Fast forward to today and I no longer have the KSA-250 or the KRC-HR. I now own an ARC REF 5SE and REF 75. I only have a scant few hours on both of them and am a long way from the magic 600 hour mark. So far, it's good to be back to tubes.
Fast forward to today and I no longer have the KSA-250 or the KRC-HR. I now own an ARC REF 5SE and REF 75. I only have a scant few hours on both of them and am a long way from the magic 600 hour mark. So far, it's good to be back to tubes.
I would advise you to try rolling the driver tubes/phase splitters in your tube amps. I had great success with old German tubes from the 50s and 60s in my preamps. There are also still some excellent US tubes around like the triple mica black plates from the fifties I use in my EARs (Sylvania, GE, RCA and others). They make a huge difference in resolution/transparency. But stocks are dwindling fast.
Stay away from JJ. I never had a good experience with them, they thicken the mids and muddy things up in general. There seem to be some OK sounding Russian Electro Harmonix around if you can't find the real deal ...
Joe's Tube Lore has a lot of good info:
http://www.audioasylum.com/scripts/d.pl?audio/faq/joes-tubes.html#6DJ8
Good luck, Jörg.
Mark
Welcome back, I understand your feeling completely , when you get use to a good thing then you will feel that should be and feel nothing new unless you make something better or go away for a while and then come back later to get those happiness again !
tony ma
Welcome back, I understand your feeling completely , when you get use to a good thing then you will feel that should be and feel nothing new unless you make something better or go away for a while and then come back later to get those happiness again !
tony ma
Mark, tubes are great and solid state is great. Today, I think the difference lies mainly in one's taste on any particular day. I do believe that ss is still slightly more reliable than tube. Particularly true IF we are talking amps with higher output and multiple tubes; not so much with tube preamps and lower powered amps with fewer tubes.
With all do respect your technical knowledge about the devices themselves, you don't seem to be very well versed in psychoacoustics. If you were, you would realize that the absolute amount of distortion a device makes is not all that relevant to what we hear. Since our ears/brains are NOT the same as an oscilloscope it is then a bit silly to state absolutist statements regarding distortion levels. What is largely relevant is pattern of that distortion and whether it falls above audible thresholds. It turns out that as long as the distortion patteren is effectively masked by ear/brain functions then it is as if the distortion does not exist. Remember the HUMAN is the ultimate detector of sound and quality...not measuring equipment. You engineers seem to forget this quite often in the false pursuit of better numbers. As to my background, I have a PhD in Analytical Chemistry. This means making meaningful measurements and harnessing technology to do so is my cup-o-tea. We battle the same issues of coming up with tests and specifications that are actually meaningful to the description of a compound. In audio, it has been known for at least 60 years that THD is a very poor indicator of sound quality... about the time that negative feedback was introduced into amplifier design.
Recently, papers by CHeever (master's thesis) and Geddes (AES conference papers) have demonstrated again that you can have very low THD and IMD and still get far inferior sound to an amp with 10x or more measured distortion. It is the pattern of that distortion that is relevant and the pattern is goverened by the design of the circuitry. Norman Crowhurst noted this as well back in the 50s that the application of negative feedback fundamentally changes the native distortion pattern and noise floor of an amp (this was back in the tube days where feedback was limited to only max about 15db...bipolars can have much much higher feedback levels). Later Otala found that back EMF can cause audible variations in an amplifier's interface with a speaker IF that amplifier has a high amount of negative feedback to be "feed" the back EMF to the input and have it reamplified.
These things affecting sound quality have NOTHING to do with the output transformer (but I will get to it in a minute) and everything to do with topology of the amplifier (also push/pull design violates the ear/brain masking pattern) and in particular NOT running Class A (objectionable zero crossing distortion), adding negative feedback (harmonic pattern and signal correlated noise floor), push/pull (cancellation of even harmonics) etc. The fact that most tube amps don't use very much feedback is probably a reason why many sound quite good and the best sounding ones use none at all...same for the best sounding transistor amps, BTW.
Now about that output transformer, it might be ringing like you claim but I have heard just as holographic 3d imaging and soundstaging in non-transformer coupled amps, albeit far less common and more importantly I have heard the complete ABSENCE of this effect in both transformer coupled tube amps and transformer coupled transistor amps (McIntosh for example doesn't have a tube-like sound at all). In the cases where it was absent I suspect that negative feedback was killing the low level information that normally would get through to create this effect. Similiarly, the OTLs I have owned had shockingly good 3d presence and soundstaging and not a transformer on the output in sight. IMO, you are taking an effect that might happen in an output transformer and painting with a very large brush.
Where the output transformer clearly colors the sound of tube gear is when they saturate. This can be seen clearly in measurements when the distortion vs. frequency is plotted. Often you will see a large increase in distortion as one gets into the bass. THis creates a lot of harmonics that go well up into the mids and gives that warm "tubey" sound that some are fond of but is distortion nonetheless. A well designed tube amp with a proper output transformer doesn't have this added "syrup". KR Audio is a very good example of a quick tight bass with a very OTL like mid and high band.
I think Lamm has hit on what Jean Hiraga and others have been saying about the distortion pattern, it matters and if you follow certain psychoacoustic principles you don't have to get rid of distortion you just need to shape it so it falls in our auditory "blind spot". The problem with most SS and even a lot of tube amps is that the distortion falls well outside of the "blind spot" and gives rise to a whole array of colorations and audible deviations.
^^ IOW engineering has to be applied to building equipment that obeys human hearing/perceptual rules rather than arbitrary numbers on paper. That is, if you want something to have good sound.
Herein lies the basis of the objectivist/subjectivist debate: We know plenty about how design affects function, but very little about how the human ear/brain system perceives sound. The more we know about the latter, the less the relevant the traditional specs on paper will become- I expect them to be supplanted by different, more accurate specs that actually relate to how we hear.
Herein lies the basis of the objectivist/subjectivist debate: We know plenty about how design affects function, but very little about how the human ear/brain system perceives sound. The more we know about the latter, the less the relevant the traditional specs on paper will become- I expect them to be supplanted by different, more accurate specs that actually relate to how we hear.
OK, so if I'm following this so far, euphonic (not the best term, I suspect) designer endeavors to create components in which the distortion is masked by an ear/brain function (This begs for further explanation. has this function been analyzed? Identified even? Are there studies? If not, how does the designer know how to manipulate it?). The objectivist - for lack of a better word - deigner endeavors to design components in which all known distortions are below the threshold of audibility?
If maskin it is a good idea, wouldn't eliminating it be a better one?
How would "masked" and inaudible sound different?
Does "blind spot" mean the distortion is inaudible? I'm very confused as to how inaudible distortion could create a characteristic sound that capitalizes on human auditory perception to be more pleasing, which is more or less what Lamm says he has done, isn't it?
It's there but it's hiding; it's not there. Is this a semantic argument?
Tim
Better words, or at least easier to understand. What are these human/perceptual rules? If they can so readily be leveraged by high-end designers, even brilliant ones, without advanced degrees and extensive clinical experience in human psychology and audiology, I've got to assume that they are simple, well-understood and broadly published?
Well, maybe not. So how do we know about these human perceptual rules, and how do humble EEs take such powerful advantage of them when the collective "we" knows so very little about them?
Tim
^^ Excellent questions!
The answer is that many designers are in fact not keeping up with the advances in human physiology. Some have not updated their knowledge base in 40 years. Some rules are well-known, others are not. To my knowledge there is no one place a person can go to brush up on them either.
Some examples of human hearing perceptual rules:
higher ordered harmonics are used to determine how loud a sound is perceived.
Louder sounds mask the presence of quieter sounds (masking rule)
Short delay echoes are used by the brain to increase awareness of the origin of the sound.
The ear/brain system perceives distortion as tonality.
We are most sensitive to sounds at birdsong frequencies.
There are tipping points where the coloration we perceive that is caused by distortion outweigh the colorations that we hear due to frequency response variation.
These examples are just a few for thought and the tip of the iceberg!
To your latter question, human perceptual rules are not taught in engineering classes for the most part; they are being discovered by neuro-science. We are very much in a situation where we are now learning this stuff; 20 years ago a lot less was known in this area. IMO/IME this is the leading edge of audio, not the development of new semiconductors or topologies.
To our current understanding, nearly the same. There do seem to be some exceptions to the masking rule, curiously enough. One exception has to do with noise that is similar to wind (hiss)- we seem to have greater perception of sounds below a noise floor composed of hiss.
I don't think 'blind spot' is entirely accurate either so maybe you have a point. Certain distortions are less audible and less objectionable so it follows that they might be more allowable in the design as opposed to others to which we are more sensitive and find more objectionable. So there is an issue of degree that is in play and it does seem to me that some designers are more conscious of such things while others are less so. I'm pretty sure we audiophiles spend a great deal of time debating the merits of these design choices
The answer is that many designers are in fact not keeping up with the advances in human physiology. Some have not updated their knowledge base in 40 years. Some rules are well-known, others are not. To my knowledge there is no one place a person can go to brush up on them either.
Some examples of human hearing perceptual rules:
higher ordered harmonics are used to determine how loud a sound is perceived.
Louder sounds mask the presence of quieter sounds (masking rule)
Short delay echoes are used by the brain to increase awareness of the origin of the sound.
The ear/brain system perceives distortion as tonality.
We are most sensitive to sounds at birdsong frequencies.
There are tipping points where the coloration we perceive that is caused by distortion outweigh the colorations that we hear due to frequency response variation.
These examples are just a few for thought and the tip of the iceberg!
To your latter question, human perceptual rules are not taught in engineering classes for the most part; they are being discovered by neuro-science. We are very much in a situation where we are now learning this stuff; 20 years ago a lot less was known in this area. IMO/IME this is the leading edge of audio, not the development of new semiconductors or topologies.
To our current understanding, nearly the same. There do seem to be some exceptions to the masking rule, curiously enough. One exception has to do with noise that is similar to wind (hiss)- we seem to have greater perception of sounds below a noise floor composed of hiss.
I don't think 'blind spot' is entirely accurate either so maybe you have a point. Certain distortions are less audible and less objectionable so it follows that they might be more allowable in the design as opposed to others to which we are more sensitive and find more objectionable. So there is an issue of degree that is in play and it does seem to me that some designers are more conscious of such things while others are less so. I'm pretty sure we audiophiles spend a great deal of time debating the merits of these design choices
Great stuff...there is a very good article in HiFi Critic from a few years ago about human hearing...and it discusses some of these points. A very simple example of human hearing (vs a microphone recording something and measuring it) is loudness. I hope I understood the article correctly. I think it said something like:
The human ear tends to compress at louder levels...so that at low levels we hear the volume actually as it is being delivered. But once the loudness increases past a certain point, our ears start to compress so that an actual doubling (3db?) actually requires more like 10db in order for the human ear to feel/perceive a doubling in loudness. The article says this physiological way of the brain/ear hearing loudness allows for the human ear to have quite a dynamic range.
So in fact, measuring the loudness in absence of this knowledge, we might say...its twice as loud because the measurements say so. but the human listener would say 'I don't hear it that way'...and would be right.
If I got that even half-right, that to me is one tiny example of where a pure measurement in the absence of understanding how the human brain/ear perceive sound could potentially lead to some incorrect conclusions or expectations in design. No doubt there are many others, as Atmasphere suggests.
The human ear tends to compress at louder levels...so that at low levels we hear the volume actually as it is being delivered. But once the loudness increases past a certain point, our ears start to compress so that an actual doubling (3db?) actually requires more like 10db in order for the human ear to feel/perceive a doubling in loudness. The article says this physiological way of the brain/ear hearing loudness allows for the human ear to have quite a dynamic range.
So in fact, measuring the loudness in absence of this knowledge, we might say...its twice as loud because the measurements say so. but the human listener would say 'I don't hear it that way'...and would be right.
If I got that even half-right, that to me is one tiny example of where a pure measurement in the absence of understanding how the human brain/ear perceive sound could potentially lead to some incorrect conclusions or expectations in design. No doubt there are many others, as Atmasphere suggests.
The problem is that there is NO SUCH THING as a linear amplification device...not tubes, not bipolar transistors, not MOSFETS or even transformers (you can get a few db gain from them but only at line level). This means that elimination is not really possible so "hiding" the distortion where the ear/brain doesn't notice it is key.
Now, you might say "Hey, I know that Halcro has 0.000001% THD so they got rid of nearly all the distortion" but what they have really done is reduced the part that isn't particularly audible and left behind the stuff that is very audible and the ear/brain looks at patterns and if those patterns are not correct it perceives the sound as unnatural. In addition, you get a signal modulated "noise" floor that is really a myriad of distortion peaks that have gone through endless loops of negative feedback. Norman Crowhurst first described this in the 1950s. I suggest reading the Master's Thesis of Daniel Cheever, the AES papers of Earl Geddes (available on his website), the theoretical simulations of Boyk and Sussman and the white paper by Nelson Pass...all of which is available on the internet by googling.
BTW, if you have heard the Halcro amp then you might understand what I am talking about that despite it's stellar numbers it sounds cold, lifeless and sterile. Not at all what music should sound like.
Assume that when I speak of eliminating distortion it's theoretical talk; I understand perfection has not been achieved. But neither that imperfection, nor all the reading you would have me do is going to give "masking" distortion a different result than reducing it to below audible levels. You can't simultaneously hide the coloration from audibility and use the coloration to create sound that triggers all the right psychoacoustic perceptions that neither you, nor Ralph, nor Lamm (need I go on?) seem to be able to explain in any way that would be particularly useful to someone voicing an amplifier.
And none of this is about those old diversions, THD and negative feedback.
You and Ralph and Lamm seem to be asserting that, using knowledge of human audio perception, audio equipment can be designed that sounds much better than it measures. I don't doubt that statement is theoretically true. But with no better explanation of what those perceptual triggers are, and how they're being manipulated by the gear in question, I have every reason to doubt that you guys are pulling it off. I can, of course, listen and hear the difference between a Lamm and a Mark Levinson. And they are different, I'm sure. Pleasant distortion or skillful manipulation of human perception to create a superior result? The latter is the claim. It has not been supported here, or anywhere I'm aware of. Is it in the papers you recommended? Time is valuable.
Tim
Last edited:
So I looked at the Boyk and Sussman paper and couldn't see where it supports your NF is evil BS, au contraire; could you point to a specific part of it?
http://www.its.caltech.edu/~musiclab/feedback-paper-acrobat.pdf
http://www.its.caltech.edu/~musiclab/feedback-paper-acrobat.pdf
No one who is trying to do SOTA gear is using coloration to give the right psychoacoustic perception. It is the FACT that certain patterns of distortion or inaudible, as long as they are not above certain thresholds, that means as long as a deisgners product adheres to that pattern (and it is one set by evolution not arbitrarily by us) it will appear to sound distortion free. In the end do you care if there is distortion if you can't hear it? The problem with techniques that are implemented to reduce distortion (namely negative feedback) is that it has the negative consequence of causing the amplifier from deviating from the right pattern, thus exposing the harmonics that are left and high order harmonics are not pleasant even to very low levels.
There is another good article by Keith Howard in Stereophile where he used a computer program to add distortion to a music file. He then compared the undistorted to the distorted files and unsurprisingly he found the undistorted file sounded the best...in other words IF the distortion is audible then it is ALWAYS a detriment to the sound and there is no such thing as "Euphonic" distortion. However, when he compared different distortion patterns, the pattern that would conform most closely to what Cheever refers to as "Aural harmonics" was the least damaging to the sound and patterns of predominantly odd order harmonics sounded the worst. Now, since we can't in the real world have the truly pure signal then the best of the worst is still the best we can get.
You need to expand your awareness if you want to claim that people don't know what's going on with human perception. How do you think MP3 was invented? They figured out how much information you could actually throw away and still get a passable representation of the original signal. It isn't quite the same but it is related. Time IS valuable and yours would be well spent reading what I have suggested and then reapproaching the subject with hopefully a deeper understanding of what's going on. The meter readers don't have this one right, I am afraid and I would argue that they set back true high fidelity (to the listener) by a good 50 years. Now, if some genius out there can come up with a linear amplification device we might make true progress to a final goal that satisfies both listeners and measurers! BTW, I am a measurer by training (PhD in Analytical Chemistry) so I know quite a bit about the limitations of measurements and the validity of specifications to actually explain a phenomenon (usually they are just good for QC work to make sure everything is working the same from sample to sample).
Then what is Lamm talking about?
No, I couldn't care less about inaudible distortion.
OK, let me make sure I get this, because if I understand what you're saying, it's the first time I've heard it
" but when we reduce those patterns of distortion using negative feedback, a negative consequence is
. So, reducing the inaudible distortions exposes the audible ones? How's that work?
Well, I agree with Mr. Howard. No such thing.
I think we've all heard that odd-order harmonic distortion sounds worse than even-order harmonic distortion. I think this is the first time I've heard that inaudible even-order harmonics can mask audible odd-order harmonics.
I'm just asking questions and looking for a solid case here. You guys may know, and be leveraging all kinds of principles of human auditory perception in your amp designs. Which ones? How? No need to reveal any proprietary technical secrets, I wouldn't understand them anyway. But I do undetstand the basics of how MP3s use perception, and masking, to work. Talk slow; I'll try to keep up.
Maybe. But, pardon my skepticism, I don't expect to find answers to my simple questiosn in the sources you refer me to if you can't or won't answer them here. Are any of these sources going to explain to me how "patterns of distortion that are inaudible" mask anything?
Tim
Tim,
Keith Howard actually goes further than that and says all distortion patterns that he tested when critically listening were not as good (preferred) as the one without distortion; however caveat maybe that we are talking about a trained listener in this circumstance *shrug*; although everyone can train themselves for distortion BUT the scope needs to be considered as preference does not necessarily has same threshold-tolerance as ABX accurate selection-identification (only other test I know done with music and specifically ABX selection-identification with THD was only accurate-statistical significant down to 3% to 4%).
Anyway it is a nice test-software-simulation he created that can be added to real music files, and AB blindly.
Cheers
Orb
Keith Howard actually goes further than that and says all distortion patterns that he tested when critically listening were not as good (preferred) as the one without distortion; however caveat maybe that we are talking about a trained listener in this circumstance *shrug*; although everyone can train themselves for distortion BUT the scope needs to be considered as preference does not necessarily has same threshold-tolerance as ABX accurate selection-identification (only other test I know done with music and specifically ABX selection-identification with THD was only accurate-statistical significant down to 3% to 4%).
Anyway it is a nice test-software-simulation he created that can be added to real music files, and AB blindly.
Cheers
Orb
We should remember that no computer program and DAC can generate the exact type of distortions induced by typical good quality analog electronics and we can not naively migrate the conclusions in this study to our audio electronics.
It is nice to read again the confirmation that people who have effective training in precision measurements of minimal quantities seem more prepared to accept the limitations of measurements and the validity of specifications to actually explain a phenomenon, using your own words - I have been working for a long time in radiation detector instrumentation and related physics.
FWIW, 'voicing an amplifier' seems like a very bad move and I don't know any designers that engage in that practice.
To understand what is going on here, the best primer is Norman Crowhurst. His texts are relatively conversational, so you can follow along without doing the math. Not to make to fine a point of it, but he was writing 50 years ago; this stuff has been known a long time.
But to give you some idea of how *we* do it: our amps are fully differential and balanced from input to output. In this way there is an even ordered harmonic cancellation that occurs with every stage, not just the output as in a lot of tube designs. In this way we avoid the most 'euphonic' of tube colorations, the 2nd harmonic. By using linear techniques (triodes, Class A operation, no transformers, single stage of gain) we eliminate the sources of distortion that many amps have to contend with; the result is we don't need negative feedback to obtain low distortion. Odd ordered harmonics do not cancel in push-pull circuits so the best thing is not to generate them in the first place. Because the amplifier has no distinct phase-splitter circuit, we don't have its attendant distortion, meaning that as we decrease the power output, distortion decreases linearly to unmeasurable. In this regard our amps, while push-pull, share the low power/low distortion aspect with SETs. There is no voice or frequency compensation of any kind, yet we get full power bandwidth from 2Hz to 200KHz (-3db at 300KHz).
The result is that the primary harmonic generation is the 3rd, but at lower levels than seen in an SET. IM distortion is kept very low due to the linearity of the circuit and the independent power supplies (including separate power transformers) for the driver and output sections. IM is also controlled by making sure that the timing constants of the amplifier do not go lower than the timing constants in the power supplies- this prevents modulation of the supplies by the amplifier, which contributes to IMD.
The result is that the amplifier is low in higher ordered harmonic generation, which means that it sounds smooth and its hard to tell how loud its really playing as artificial loudness cues to the human ear/brain system are minimized. So this imparts a relaxed presentation but with obvious speed and detail at the same time. The weakness of course it that we can't play the amp on just any loudspeaker, but what we found out decades ago is that if the speaker demands that the amp use feedback to sound right on that speaker, the chances are remote that the speaker will sound like real music with any amplifier. IOW we limited our marketplace to only those speakers that have a chance of sounding like real music. Fortunately there are a lot of them out there to choose from.
FWIW its the lower ordered harmonics (2nd, 3rd, 4th) that are considered musical to the human ear. Audiophiles use terms like 'richness', 'warmth', 'lushness' to describe them. Regarding your final question, I don't know. But I can tell you how it works:
Per Norman Crowhurst, the use of loop feedback adds additional odd ordered harmonics that did not exist prior, up to the 81st. In addition, you get intermodulations caused by non-linearities at the feedback node (the point where the feedback mixes with the input signal). Essentially this becomes the noise floor of the amplifier (and is the 'harmonic spray' that Tomelex likes to talk about).
Now if you recall the masking principle: The ear cannot hear anything in the signal that might be below this harmonic noise floor of the amp. However, if you don't have any feedback the noise floor of the amplifier will be mostly hiss. It appears that the ear can hear into that kind of noise floor (I have a theory that hiss sounds like the wind, and for survival/selection purposes our ancestors evolved a means of hearing some detail that exists below this type of noise floor) and so if you don't have feedback, you can extract more low level detail. I believe this to be the answer to your question.
- Status
