 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
98.144.114.174
In a past life, I owned many tube amps. Almost all of them were single-ended, and almost all of those were DH-SET: 45, 50, 2A3, 300B, and - especially - the big transmitter triodes (211, 845, 805).
I largely stepped away from the hobby, and was very happy listening to my old, restored Quad 63/988 ESLs with a Schitt Yggy DAC, Schiit preamp, biamping with an active crossover to hefty subs, and a Purifi-based amp powering the Quads.
This lasted several years.
I think I fell for the left-vs.-right-brained listening mode tricks again because, on a whim, I tried an $850 300B amp from Amazon in place of the Purifi and have had a few of those "I'm hearing everything new again" sessions.
It's not "I like euphonic distortion." It's not "I like soft clipping." The reason I know is that what is wowing me is the low-level stuff: The extra texture I am hearing in all kinds of things (it's not "uniform" so it can't be some intrinsic coloration) and the extra nuance and detail. This goes all the way from Depeche Mode to Patricia Barber to Freddie Hubbard.
(I learned these things about SET amps long ago, but have learned them anew.)
HERE IS THE PURPOSE OF THIS POST: I am not looking for any kind of validation of this experience (subjective as it is), or anything of the sort. The question on mind now is *what is responsible for this*?
The Purifi amp is perfect in every way that can be measured. I am entirely aware that measurements are limited - we don't measure amplifiers playing *music* - but, still, an amp that has THD and IMD distortion lower than the threshold of the measuring equipment, at all power levels (a first), is, at least, a very good performer.
I am focusing on two things now: No splitting of the waveform, and transformer coupling to the speaker. Is the secret one or both of these things?
From what I know of digital amps, it seems possible that crossover distortion is nonexistent. But I'm not sure. And what I am hearing now from this 300B tube indicates otherwise, frankly, because I don't know how else to explain this, except that the SE amp is preserving tiny signal levels better.
Or is it that output transformer? I read something, once, about the transformer insulting the amplifier from the back-EMF of the speaker having very good effects. I don't know - and I don't know how that applies to an electrostatic, which is a very different type of load.
It's often pointed out that an audiophile's power amplifier is the very last stop for a signal that has been through countless steps of manipulation and amplification; how can this last step somehow be so "special," have such a large effect on the actual sound?
I guess the answer is twofold:
- It's by far the largest magnitude of amplification
- It involves interaction with the transducer
I am looking for people with greater knowledge of audio engineering than I to chime in here. Describe what I am hearing.
(Just don't tell me that I'm *not* hearing greater low-level nuance, because reality is reality.)
Follow Ups:
Tonight I listened to Depeche Mode's "101" with my Chinese 300B integrated powering my Quad 988s (actively crossed at 90 Hz this evening) and heard things I have never heard before in this recording.
Many things.
Good things.
Not bad things.
New data point: My DAL Audio LM3875 Gainclone Monoblocks
I dusted them off today (literally - after a few years - and I will never sell these amps) and found that they are *much* closer to the SET sound than the Purifi amp.
They have tons of texture and tonal color. They are extremely fun to listen to.
The Purifi amp surely has a lower noise floor and more "detail" in some respects, but... after listening to the 300B SET and now the gainclones, I realize why I'd stopped listening to acoustic music the past couple years.
Where the cheap, $850 Chinese 300B integrated leaps above the gainclones is in treble refinement. It *never* sounds uncouth - and if it's "rolling off" anything, I can't hear it.
(I'm going back to TX coupling with the speaker as the main explanation for SET/tube sound. Apologies for lumping those categories.)
-being a 300b amp, you might consider trying out a PP amp of the same power and see what you think.
The tricky bit when people compare SETs to other tube amps is the tube types used, the maximum power available, and the parts used to built them.
The output tubes are almost never the same. Neither is the power level, and with tube amps, SET or PP, the lower the power the better the output transformer can be. Especially when there's no feedback, these things can be tremendous variables.
FWIW:
I know of a 5 Watt PP amp that challenges SETs of any cost.
If you have a Dynaco ST35 (the smallest amp they made) properly fixed up it can be remarkably convincing. IMO you get much bigger than that and the 'magic' of tubes is far less profound.
This is all about 'how much better can it get', right?
I've never preferred any P-P TX-coupled tube amp (of any type) to a SE (of any type) in the past.
That made only 7 Watts?
All of them
For example, we make a PP amp that makes 5 Watts. I know you didn't use that one...
More seriously the real question is 'how much better can it get?'
To sort out what SETs do and what they don't do, the first thing, IMO/IME is to compare them to other amps. To do that with PP tube amps you really need to level the playing field.
PushPull tube amps have output transformers too and to a lesser degree they are sensitive to how much power the transformer is asked to pass. SET output transformers are really sensitive; PP outputs a bit less so.
So to level the playing field so you can hear what the difference really is, you have to do one or more of several things.
You could have the PP amp use the same power tubes as the SET. Or you could simply have the PP amp be rated the same power. Its also helpful to use the same quality of parts and construction technique.
I've found those things are difficult! Most PP amps make a lot more power; PP amps based on DHY (Directly Heated Triodes) are rare. Low power PP amps of good quality are rare too. This caused me to build both types.
What I found is SETs have no musical advantage at all. If you have a competent PP amp of the same power or using the same power tubes, the so-called advantage of SETs is simply erased. IOW, the 'magic' is still there in lower powered tube amps if they are built well.
But it gets better. There are class D amps that can take on SETs just as easily. I really don't think there's any advantage to SETs at all and the reasons they went away 70-80 years ago are just as valid today as they were back then by any metric audiophiles value.
So IMO this is really just about the small sample size SET lovers have heard and nothing else.
...can be made by using whatever active device (tubes, discrete SS, integrated circuits), operating class (A,B, AB, D), either SE or PP, with or without negative feedback.
Only the corresponding recipes (and cost) differ.
Omnes feriunt, ultima necat.
I was hoping you'd chime in, Ralph.
But, you are the guy that taught me that any TX-coupled PP amp suffers hysteresis loss in the transformers, robbing detail.
And, also, going PP with tubes for sure means crossover distortion.
Don't these things contradict your assertion pretty seriously?
(These are things I seem to hear, as well, with PP tube amps, even really expensive ones.)
But, you are the guy that taught me that any TX-coupled PP amp suffers hysteresis loss in the transformers, robbing detail.
And, also, going PP with tubes for sure means crossover distortion.
Don't these things contradict your assertion pretty seriously?
(These are things I seem to hear, as well, with PP tube amps, even really expensive ones.)
Yes, all audio transformers introduce distortion due to hysteresis.
However your comment about crossover distortion is simply false. This is one of those things you probably should confirm by measurement, once you've identified the amplifier to which you make this assignment.
I've not found a PushPull amp that makes crossover distortion, although I've read of a few- I think there's one by Douglas Self (called the 'blameless amplifier' which is a class C design so it has crossover distortion problems). BTW, the nonesense spouted by morricab is fictional. He distorts things I said online and he incorrectly applies the writings of established authors in his posts, such as the one below, in a way that is misleading.
Despite his posts being proven incorrect, he's stuck to his position; my theory is due to the fact that he sells SETs for a living. Sometimes the truth is inconvenient.
I'm not attacking SETs so much as I'm simply pointing out their weaknesses; once you know them you are in a better position to get the most out of them. I can do that for a lot of amplifiers and have done so elsewhere. Nothing is perfect; getting the most out of any product is just good common sense.
Douglas Self's amp is not class C Ralph.
-
In recent shenanigans on this forum about bias classes I noted that Self preferred something he called optimal class-B where bias is adjusted to give the smallest perturbation as you cross between pull and push but I didn't know what that equated to in terms of bias current (for a complimentary emitter follower stage). My interest reignited I obtained a copy of his Power Amp design book and have been working my way through it. As chance would have it, today at lunch I was at a section where he showed such numbers and an optimal bias turns out to be in the 100mA to 200mA range. He finds the optimal bias should set by the voltage across the emitter resistors and not by the current through the followers (and the optimally biased current is much lower when using complimentary feedback pairs instead of followers). What is very interesting is that he shows how the output stage small signal gain changes from under-biased to class AB along with the corresponding distortion residuals.
In Self's defense I think it fair to say he does know the difference between bias classes and he shows that there will always be cross-over distortion in a solid state push-pull output stage even if it is very small.
I know little about tubes but I find it difficult to believe that turning one off as part of a push-pull hand over does not create some discontinuity and I boldly, yet respectfully suggest that if you haven't seen it you are not looking hard enough.
I'd like to find a tube amp with that problem, brought on by bias in the output section (as opposed to a driver problem). I've been working on consumer electronics since the day after I graduated high school in 1974 and worked my way through my university years doing that, and for years after until I got Atma-Sphere off the ground. If you can point me to one I'd be interested.
FWIW in the hifi world the tube amps with the least bias current I've seen were made by Electro Voice. They described them as class B. I have one and while it does make double the power you'd expect from a pair of 6V6s (20 Watts) its actually a very lightly biased AB design. At any output level you can measure there's no evidence of any discontinuity with sine wave reproduction.
Class C is defined as 'less than half of the waveform'. I think we all know that, but really haven't thought of the implication.
Class B is exactly half. You can see that in practice, this is difficult to actually achieve, although today with microcontrollers you might be able to pull it off. IOW if the device is cutting off a 1/2 mV prior to zero crossing, that's by definition class C. IOW I doubt any class B amps actually exist- they are some form of AB (as the EV amp above) or they are C.
I know Self knows the definitions, but I suspect he really didn't think it through. I know I didn't!
Let's have a moratorium on class definitions. We know what they are, instead give some thought to the continuum between them. As noted in other parts of this thread I don't know about tubes but I don't imagine that the transfer curves are so radically different from solid state that there will be no cross-over discontinuity.
So, I went to Stereophile and looked at the measurements of several tube amps as JA does notch out the fundamental and show distortion residuals. And all the tube amp residuals are nice third harmonic sinusoids with a bit of second thrown in. The THD for these residuals is in the 0.X% range so any spikeyness from cross over is probably buried. Compared to the the Boulder solid state amp (discussed in a thread above this one) the distortion residual is noise-like but you can see distinct steps that must be associated with a device turning off. Also note these don't line up with the zero crossing and this must because this is a classAB design that gives you two cross-over discontinuities as you leave Class A to class B twice per cycle, one on the positive side and one on the negative side.
Edits: 05/24/25
I've serviced many solid state amps. If they ever exhibited crossover distortion artifact I would turn up the bias until they didn't. Usually that was not that much. If the amp is AB I would not expect troubles with that plain and simple. But if the amp is biased with less or no current at all in the output section at idle I would expect problems.The thing about class B is its really a thought experiment, or was until the advent of micro controllers. To be class B you have to cut off at exactly zero, nor more or less. If less than B its C (only push pull; expect crossover distortion issues) or its class AB, so its biased enough where that's not a problem. However you can still have crossover distortion problems if the driver lacks the current to deal with the input capacitance of the output devices. I regard that as a design problem rather than a class of operation problem.
Edits: 05/27/25
Isn't the simple fact that the two (at least) output devices handling the signal will never have exactly the same operating characteristics (curves) enough to cause some kind of distortion? Would this not be "crossover distortion"?
Care to comment on this?
https://benchmarkmedia.com/blogs/application_notes/131424519-amplifier-crossover-distortion
https://benchmarkmedia.com/blogs/application_notes/131919495-ahb2-crossover-distortion-measurements
If the two devices have slightly different characteristics (and they all do) they will conduct differently. So if they are biased such that neither goes into cutoff, what will happen is a bit of lower ordered harmonic will be created as the signal level goes up.
The article at the first link describes a class C push-pull amplifier in the opening remarks. It does not describe a class A or AB amplifier. You can clearly see the waveform drawn to show where the output devices have clearly gone into cutoff just shy of the zero crossing.
In an AB amplifier this simply does not happen. So you don't get crossover distortion.
I am skeptical of the article at the second link. The reason is simple: in both cases a sine wave is displayed. This means (assuming feedback is used) that the feedback is able to control distortion at the zero crossing point.
But if the devices have gone into cutoff, there will be no feedback so there would be a zero crossing artifact; IOW with no feedback the amp would be unable to correct at that point.
So this casts doubt on the measurements as shown.
The 'competitor's amp' moniker is also less than convincing... IMO this is an example of marketing.
So, to be clear, you are saying that this statement in the article is simply false?
"Note that crossover distortion is clearly visible as sharp transients in the distortion waveform. The transient shows a fast attack time followed by the recovery time of the feedback loop. This distortion waveform is typical of virtually all class-AB amplifiers."
For example the amount of distortion isn't shown.
We see an alleged 1 Watt waveform, which seems to be pretty nice. Go back and look at the first link you posted though- there's a waveform shown there too.
Those two don't look the same. I've seen such waveforms though when solid state amps didn't have proper bias adjustment. How do you square the difference?? On the one hand we have an obvious distortion problem and on the other, despite a clean sine wave, a claim of distortion.
There are too many variables not nailed down at the second link, so I'm taking it with a few grains of salt.
I think your points are valid.
I'm surprised others aren't calling out Benchmark for this. These amplifiers are wildly popular.
Where is Ralph wrong, in particular? I'm listening.
He is wrong in saying that Class AB and Class B amps don't have zero crossing distortion as Benchmark rightly claims. Only Class A amps are truly free from this distortion. Benchmark claims they found another way to eliminate it with feed forward...- this I cannot say whether it works or not.
I meant in the specifics of the critique of that Benchmark article.
It seemed to me he had a point regarding the sine wave images.
There has been some discussion recently on this forum regarding bias classes and cross-over distortion. I have written before that solid-state follower power amp classes are really a continuum that deserve more consideration of what is happening than effort expended in classification.
And I've written that cross-over distortion exists in all follower stages except class-A. A good resource is Douglass Self's Power Amplifier design book and, to help draw a line under this topic, I have captured some of the figures to reiterate these points (colored notations are mine).
Below is the simulated small signal gain of a BJT complimentary follower stage. Ideally the voltage gain would be unity but with loses it is less than unity. The x-axis is the output voltage, going from negative on the left side (the PNP pulling the load down) to positive on the right (the NPN pulling the load up). Various bias settings are shown and note that none have a constant gain as the output voltage traverses the range.
Self gives a reference that proves it is impossible for complimentary BJTs to have constant gain across the hand-over and the simulation shows it. Any wiggle in a transfer curve equates to distortion and the sharper the wiggle the higher the harmonics that will be generated (think of fitting a power series to a line, the sharper the curves the higher the order of the fitting polynomial). The green circle is Self's 'optimal class-B bias point' chosen for the least wiggling (NOTE, as I post this I think the green circle should be on the curve above). If the bias is reduced then the small signal gain takes a precipitous dive downwards and once the bias is low enough to reach a gain of zero that would represent class-B with the devices turned off. Why anyone would even consider this, I do not know. Going the other way, applying more bias moves us into class-AB but that does not make the overall transfer function any smoother. It is more continuous at zero but now we're developing more serious wiggles where the transistors do turn off around +/-5V for this case. Extending the bias into class-A gives the plot below:
Here we see smoothness across the zero crossing with gain droop at the maximum voltages where a device is starting the turn off at the limit of class-A current. The continuity from under-biased class-B through 'optimal class-B' to class-AB to class-A is now apparent.
Let's now have a look at the cross-over distortion created by the transfer curve wiggling. Below is a distortion residual from one of Self's amps that has most distortion mechanisms minimized leaving cross-over distortion apparent in the residual. Note this measurement was made with a lot of averaging and the amount of sharpness captured will depend on the resolution bandwidth of the oscilloscope. Note also that this distortion residual is 0.0004% of 25W into 8ohms. Compare this to the distortion residuals of tube amps shown in Stereophile measurements - there the THD is in the 0.X% range and the residuals are usually a fairly clean third harmonic swamping out any cross-over.
Here is an under-biased case as we head down that gain null creating a larger, sharper wiggle in the transfer curve. The zero crossing spikes are now much larger (though there is no note about their amplitude compared to the previous graph or the corresponding THD)
And, lastly, the distortion residual for a class-AB case
Note here that sharp changes in the residual are not aligned with zero crossing but now there is one one each side of the zero crossing for each transistor turn-off one side of the zero crossing
In summary, we should expect cross-over distortion in anything but class-A even if we have to work hard to capture it as it might be small. Though these pictures are for solid-state follower amplifiers, I can't image a push-pull tube design can be immune to cross-over distortion, though it might be buried by large low-order harmonics.
As for the Benchmark material, looking beyond the sales & marketing hurrahs I don't see anything obviously fallacious there. The distortion residuals in their plots are amplified by 60dB compared to the output sinusoids. For the competitor's amp the residual is of similar size so it must at least -60dB or 0.1%, roughly (they don't quote the competitor's 1W THD). The AHB2 quoted 1W distortion is 0.0004% so that is 250x smaller. It seems not unreasonable to me that if you shrank the competitor residual by 250x it would look like the AHB2 straight line on the same scale.
If the device does not go into cutoff its class AB of some sort.
If it cuts off prior to reaching zero, by definition its class C.
Class B is really theoretical; I don't think anyone has ever made one. The Self amp appears to be class AB, with the outputs biased very low resulting in poor linearity near zero crossing. So there is a bit of distortion generated as a result.
To be clear: class C is defined as 'less than 180 degrees of the waveform'. A class C push pull amplifier is possible; the output devices might conduct for only 179.5 degrees, cutting off just before zero.
Put another way, so-called 'class B' amps are actually either class AB or class C, unless one chooses to ignore the definition of what class B really is.
No disagreement from me with anything you put in this post really. It is what I have been saying across several threads on this forum while getting shouted down by Ralph that I don't know what I am talking about.
I told him clearly that ONLY Class A will not have crossover distortion and that Class B and even a lot of Class AB will have visible zero crossing distortion. He then tried to deflect and say that those must be Class C, which is nonsense.
We can debate the audibility of this until the cows come home. Suffice to say that I think it is nearly always audible and serves to destroy the realism in push pull amplifiers (both SS and Tube) that are not true Class A.
The ONLY PP amps that I have heard (tube or SS) that I find not objectionable after some time are Class A biased amps, like my friend's Plinius SA103, which has both A and AB modes. In A it sounds pretty nice, although that odd order dominated spectrum still renders it a bit grey sounding compared to the best amps. PP triodes in Class A can sound very good. For whatever reason, the cancellation of even order harmonics is often not as complete for tubes as it is for transistors, leaving some psychoacoustic benefit I think.
This for me leaves only single ended amps (tube, transistor or a mix of technologies) as having the most natural sound...they may not be ideal for all speakers...but there are now sufficient numbers of easy to drive speakers that I need not worry about speaker choice.
There is nothing to call out. Benchmark is correct and Ralph is wrong. While I am not a fan of their products, they are clearly very good engineers and know their stuff.
I posted many measurements from Stereophile of recent AB amps that show exactly what Benchmark is talking about.
Where'd you post that stuff - the Stereophile measurements? Thanks.
Look in threads down below...
-
Makes sense.
So hysteresis loss occurs even in SE circuits? I may be recalling wrongly, but I could have sworn that when I read that post from you (years ago), you were talking about PP circuits, the reason for the loss being crossing the 0-point which doesn't occur in SE.
Please clarify/educate me.
I don't measure amps. I'm not (really - I've built a few kits) a builder or manufacturer. But, I have seen crossover distortion measurements from plenty of PP amps - although SS, non-TX-coupled, comes to mind. For example, both Hegel and Benchmarch cite these measurements in their respective discussions about how their proprietary tech eliminates it.
How would crossover distortion be eliminated entirely in a PP circuit unless the two (or more) devices had absolutely identical operating properties around the crossover point? (I do understand how bias effects this, I think - but that doesn't eliminate this problem.)
So hysteresis loss occurs even in SE circuits?
Yes. But in a different way from what happens in a PushPull circuit. In an SET the loss is less because the sigmoid curve is in only one quadrant.
I have seen crossover distortion measurements from plenty of PP amps - although SS, non-TX-coupled, comes to mind. For example, both Hegel and Benchmarch cite these measurements in their respective discussions about how their proprietary tech eliminates it.
How would crossover distortion be eliminated entirely in a PP circuit unless the two (or more) devices had absolutely identical operating properties around the crossover point? (I do understand how bias effects this, I think - but that doesn't eliminate this problem.)
If a solid state amp exhibits crossover distortion its mostly caused by the bias not being set up correctly if the amp is intended to be class AB.
Its not hard to make sure crossover distortion isn't a problem. Morricab and I went 'round and 'round on that one a few weeks back on the thread below.
Actually since then I sorted out that class B amps don't actually exist; class B exists in theory only. Real amps are either AB and very lightly biased or they are class C (and so have problems with crossover distortion, although there are ways around it as I indicated in the thread below). If you know the definition of a class C amp then you know what I'm talking about.
To make sure you don't have crossover distortion the output section must be biased enough the output devices don't cut off at zero crossing. You must also have a driver circuit capable of driving the input capacitances of the devices at all power levels. If the driver runs out of gas you can have a crossover distortion problem that has nothing to do with the class of operation.
Ralph has been waging a war on several forums against SET and those who think SET sounds better.
I, like you, have tried to move away from SET in the past only to rediscover the simple fact that they sound more like the real thing than other amp types.
I have done exactly what Ralph has prescribed by owning a couple of lower powered Class A PP triode amps (VAC 30/30, PureSound A30, WLM (made by Trafomatic) Class A EL84...etc.) and listening extensively to very good Class D (Mola Mola, Devialet etc.). None of them achieve what a good SET achieves.
The Class A triode gets pretty close because in Class A there is no crossover distortion; however, a push/pull circuit tends to cancel out even harmonics and this means the amps do not adhere to what Daniel Cheever calls the Aural harmonics, which Jean Hiraga also noted. You need an exponential decay of harmonics, starting with the 2nd harmonic, where each even harmonic is higher in amplitude than the following odd harmonic.
Ralph's OTLs and for sure his Class D amp do not follow this prescription for good sound because they have odd harmonics as dominant. This was noted by audio journalist Keith Howard as being the most offensive harmonic pattern of those he tried when he added distortion digitally to audio files. Of course he found the unadulterated to be the best but the least offensive was an even/odd pattern with the evens being higher in amplitude than the following odd.
Cheever found this also to be true when he developed a metric for evaluating amplifier quality. It is related to the ear's own self-distortion that the brain has evolved to ignore. That pattern is an exponential decay of even and odd harmonics. He posits that as long as the even/odd distortion of the amp is in the same pattern as the ear's own pattern and lower in level than that ear generated distortion then the distortion from the amp "hides" behind the ear's own distortion and the brain ignores it. The amp then sounds completely (or nearly) distortion free. This is consistent with audio journalist Jean Hiraga's assertions about a generation earlier.
It also fits with earlier attempts to develop meterics from the likes of D.E.L Shorter (BBC engineer) where he "weighted" the increasing harmonics in his equation to give greater importance to the impact on sound the higher the harmonic order was.
Ralph talks a lot about the SET limitations and the criticality of the output transformers. He is not wrong in the sense it is a critical component to getting a good sounding SET...so is the driver stage. Most SETs, especially those for reasonable prices, will necessarily skimp on the output and/or interstage transformers because they are the most expensive bits of the amp usually. An undersized transformer will make issues with bass saturation, which gives a significant coloration to the tube amp right up through the midrange. Really good SETs will have a properly sized transformer and the necessary winding complexity to allow good high frequency extension. Even very large core transformers can be wound in such a way to provide good extension.
The driver tube and mode of drive is also critical. I have found by owning a lot of SETs that I tend to prefer ones that are transformer coupled rather than cap coupled or direct coupled. The reason is dynamic contrast. They feel as if they have more macro and micro dynamics and simply more grunt. This was true of my 25 watt Aries Cerat Genus and Diana integrateds as well as my 3.5 watt Silvercore 2A3 amp. The SETs I have been most disappointed with were cap coupled using a 6SN7 as a driver tube, of which I have had a few.
Push pull amps can only completely avoid crossover distortion if they are running in full class A. If you look below there is a long running argument over this and I showed that there are a lot of modern Class AB amps that still visibly exhibit this very nasty type of distortion. I am pretty convinced that this is one of the big culprits in why a lot of amps sound simply synthetic and lack natural tone and body.
The other is negative feedback and interactions with the speakers. Your Quads will pump nearly 100% of the signal back into the output stage of the amp. If there is high negative feedback, then there is a pathway from the output back to the input, which is then reamplified. Of course what comes back out of the Quad looks nothing like what went into it.
The VAC 30/30 that I owned had an interesting feature on that allowed adjustment of the negative feedback from 0 to 8 dB. It invariably sounded the most natural on 0dB. Other amps that have had that feature in the past were also reported in reviews to sound the best either on 0 or on the minimum setting allowed.
I think the solid state amps from CH precision allow feedback adjustments as well and most prefer the minimum settings.
The discussions from Nelson Pass also indicate that he is an advocate of finding linearity without the use of negative feedback.
Ralph has now gone 180 from his OTL past into the Bruno Putseys Class D and ultralow distortion camp. I am not sure what to make of this but a lot of his anti-SET rhetoric on this and other forums seems like a large conflict of interest given he is now selling Class D amps, which are kind of the antithesis of SET.
I'm familiar with the distortion spectra research, of course. I don't doubt that it helps to have even spectra present, but I do not think this is important to SET sound, or at least not the SE sound characteristics I find appealing, because we're talking about very low-level signals, where all THD is miniscule (even with a SET).
Rather, I think monatomic distortion - decreasing linearly with level - is relevant. But, then, I'm pretty sure the switching amps do this too.
I think the Atma-Sphere switching amps sound extremely good. I've heard them once, in what I thought was one of the very best rooms at Axpona 2025, with Popuri electrostats.
Then again I think my Purifi amp sounds very good as well.
After more listening, while I find the SET sound more *enjoyable,* I can't be certain it's more *accurate.*
I am leaning more towards the transformer insulation effect being responsible for this SET sound - but, of course, that should then work for PP TX-coupled tube amps as well, and it does, pretty much, but not with the same level of low-level naturalness, in all of my experience. I've always attributed that difference to crossover distortion.
I've always attributed that difference to crossover distortion.I don't think it is.
Norman Crowhurst pointed out a good 60 years ago that a PP amp with a single-ended input (like a Dynaco ST70) would have a bit of an emphasized 5th harmonic, due to the combination of quadratic and cubic non-linearities.
When most people compare SETs to PP, its exactly that sort of PP amp that is compared. But if you put such amps (like the ST70) on the bench you see there's no crossover artifact.
I have an old EV amp that is claimed to be 'class B' (which, when you really think about it, can't exist easily in the real world; more likely some kind of AB circuit, if not the its actually class C...) and it exhibits no crossover artifact at any power level as low as can be measured.
So Occam's Razor suggests a simpler explanation: its something else. I think its that 5th harmonic not being masked properly.
To get around that 5th harmonic problem the PP amp would have to be entirely balanced or entirely differential; perhaps both. In that way the distortion is more the result of a cubic function, which is IMO vastly superior to a quadratic non-linearity insofar as the human ear is concerned.
Edits: 05/22/25
Keith Howard thought so too...until he added small amounts of distortion to digital files... was clearly audible and a clear ranking of preference was determined.
Just a question, why wouldn't more enjoyable be more accurate? Isn't music supposed to be enjoyable? Or do you think most recordings are very poor and therefore would sound not very enjoyable with more "accurate" amps?
Not sure that switching amps do this...at least the earlier ones definitely did not.
My test is simple. Use the most sensitive speaker you can find, which basically acts as a magnifying glass on any distortion artifacts, and if it sounds really good with that then it is likely to be low in PERCEIVED distortion, regardless of the measurements.
To date, I have not heard a single SS or Class D amp that sounded natural on a horn...not for lack of myself or others trying. Usually, it is like fingernails on a chalkboard. This years Avantgarde demo in Munich is just another data point in that observation. It was horrible and aggressive with their built-in Itron amps. As AG is one of the most sensitive speakers on the market, I think it is a fair speaker brand to base this assessment on as they are present also at most audio shows here in Europe at least. I have heard them with a wide range of amps but by far the best were the Thomas Mayer tubes (SET amps) and the Audiopax amps (SET with KT88) from Brazil. A PP tube integrated amp from Octave (V40) was ok as well. Every time I have heard them with SS it is a disaster.
Your explanation for SET sound being derived from insulation effects seems like a strange argument. On the one hand you claim that the distortion, even of most SETs under normal conditions, is too low to be having the claimed effect and then you bring up tranformer insulation, which must be orders of magnitude lower in distortion, as a likely culprit. Is that logical? I mean, it could have an effect but that would also assume that all transformers have the same insulating materials and I am not at all sure this is true.
First, it seems none of us know what effect transformer isolation has on an amp. Have you measured an amp, playing music signals while also interacting with a real (speaker) load? Nope - nobody has. We don't have the tools to do that, really, I think.On another note, there are some big problems with your amp-quality-assessment methodology above.
First, high-sensitivity speakers often have serious FR aberrations - both rising response and peakiness. It may be that an amp that is more accurate in terms of FR response sounds worse on such a speaker, completely independent of distortion.
Relatedly, of course TX-coupled amps behave very differently into real loads than do non, which means more frequency-response differences.
Horns: I've had a lot, front and back. Probably the most revealing, certainly the most esoteric, were Cogent field-coil compression drivers in conical horns. (I had only the mid horns, mated with BR bins.) On these horns LM3875 gainclone amps were highly competitive with premium low-watt SETs (45/50 tubes). OTOH at the time I know I was not listening for the things I find most appealing now with TX-coupling - and also there was really no treble in this system.
Edits: 05/28/25
The transformers sound different on two accounts, the sigmoid stays in the first quadrant, unlike PP, and there's a cut core, which limits low frequencies.
We seem to be on the same page WRT horns.
FWIW I've been using horns for years with class D and its quite relaxed. The class of operation and topology does not matter as long as the distortion signature is benign. Jean Hiraga reviewed one of our OTLs about 25 years ago and seemed to like it quite a lot; he gave it 9 or 10 out of 10 for every metric.
Your fictional stuff above suggests you could be good at it.
Oh you mean like your misinformation about not seeing any PP amps with crossover distortion when I gave you a list of several recent designs from Stereophile measurements in a thread below?
Or you misrepresenting Douglas Self's amp as Class C?
and then class B. If you know the definition of either, you'll see I'm correct.
See the Dukes post above. If you can read Self's work then you know he knows what the Class of amp operations are...
I'm in a similar situation, I had a beautiful set of audio note Neiro's with 2A3's. They did the " you are there" thing. Equating 2nd harmonics with resolving low level detail does not make any sense to me and a well designed amp into efficient speakers will have very low distortion, this is a low hanging fruit explanation, I do not see how even order harmonics would reproduce a pin dropping in the 12 row better than no distortion and if that is the case I want a distortion knob on my next amp. I don't have an answer but those who know, know.
It has to do with what your brain does with those low order harmonics. They are being masked because your ear makes similar distortion patterns. This means that what you are hearing, as long as you don't push the amp too hard, is effectively pure because your brain filters the harmonics out. Neat trick but it only works if the pattern of the amp mimics the pattern of the ear's own distortion.
I have heard a manufacturer say such a thing and it seems plausible at first instance but I think the claim doesn't stand up to a little thought. That the ear creates distortion I have no doubt - the eardrum membrane pushing against a little bone that pushes against another little bone that pushes against a third little bone that pushes against another membrane (transforming a small movement of the large ear drum to a large movement of the small opening to the cochlear) that vibrates fluid in the cochlear that in turn vibrates the basilar membrane that moves the cilia that cause nerves to fire...etc. Not audiophile designed. But, surely, your brains knows how to distinguish between sounds originating from outside the ear to those created inside the ear and would only cancel what is creating inside? Everything we hear is distortion, the difference between a piano and a violin playing the same note, the character of a voice is all distortion, we never hear pure tones. And the only way to test this is to strap headphones onto a test subject and ask them to compare signals so all you are testing is the threshold of audibility of second harmonic not why the threshold is what it is. But, there is a ton of research by hearing psychologists so maybe someone has studied this but I kind of think they are not too inclined to placate audiophiles.
In similar vein, Ralph brings up the lower order distortion masking higher order distortion. This is plausible too but I would love to see this quantified with some examples to see if it really is a thing and compare it to the masking effect of the fundamental itself. As you consider higher harmonics the frequency separation from them to the second and the fundamental become closer and the amplitude of the fundamental will be much larger than the second harmonic. Is masking of high harmonics by low harmonics real or just a conspiracy by tube amplifier designers put about to sell tube amps? :)
If you don't agree, ok, however, it seems quite plausible to me.
https://next-tube.com/articles/Cheever/cheever.pdf
Instruments and others voices have significantly different patterns from what your ear would make. How do we know what the brain is filtering and what it is not, other than whatever overlay of harmonics the ear makes on incoming sounds is not perceived in and of itself.
I think it explains the resurgence of SET better than any other story I have heard so far. It is also consistent with what renowned audio journalist Jean Hiraga came up with a couple of decades earlier regarding what distortion patterns he thought gave the best sound.
https://www.stereophile.com/reference/406howard/index.html
This experiment by Keith Howard is also very interesting. He wrote some code to add distortion to a digital audio file. He experimented with patterns and levels and found, unsurpisingly, that the unadulterated file sounded the best followed by the even/odd exponential decay pattern. The worst was the all odd pattern (typical of push/pull amps because they cancel to a large extent the even harmonics).
My own anecdotal evidence is also in play, every audiophile that I am in contact with has either SET, TriodeFET (single stage single ended hybrid), PP Class A triode or Class A PP SS. No one has Class AB or D and most are single ended. ALL of them came from high power, low distortion Class AB or D amps...including me. Definitive? Of course not; however, it is telling that as they gain more experience this is the direction they drift.
1) it was well-known by the 1930s at least that the harmonic spectrum was important. I refer you to the 3rd edition of the Radiotron Designer's Handbook. In it we see that it was well known that higher ordered harmonics were unpleasant. I didn't see Cheever mention this and it has bearing on his conclusions.
2) Crowhurst pointed out and Baxandall 15 years later that if the feedback signal becomes distorted by passing through the cathode of a tube or the base of a transistor (in the latter's case) that the result will be distortion. Crowhurst did not propose a solution; Baxandall merely proposed just using more feedback.
There is another method which is to use the means that opamps use; mixing the feedback in a Voltage divider network rather than in an active non-linear device. If you follow that practice you'll find feedback working considerably better.
3) Bruno Putzeys explains some of the rather silly stuff that was going on the 1970s in his paper the 'F Word'. You and Cheever should read it. I'd not be surprised if Cheever may have already.
4) On that note, no mention is made of what happens when distortion rises with frequency. See Bruno's paper about that. Zero feedback amps are capable of no distortion rise across the audio band. Most older amp designs using feedback are not. If you want to talk about how higher ordered harmonics can affect the presentation, this can't be ignored.
5) on page 31 we see 'it sounds more dynamic'. This is the direct result of higher ordered harmonics on the leading edges of transients, where more power is required. Its an illusion , one in which Cheever was taken in. A sound level pressure meter reveals the truth. Not more dynamic, simply more distorted. This is very easy to demonstrate with simple test equipment; I have to admit I was surprised he was so easily taken in.
6) the idea of the ear's self generated harmonics, while real, WRT to how we hear differences in amplifiers is so much poppycock. He really should have thought that bit through a bit more carefully. By his metric we'd be unable to tell the differences between instruments made by different artisans. Yet we do; the result of millions of years of evolution, which cannot be ignored. Put simply our ear/brain system has developed to deal with internal distortions and does pick out distortions caused by amps, including the 2nd and 3rd harmonics, no matter if a proper exponential decay exists or not. He just didn't think that bit through.
While I agree with him that the current measurement regime fails in that it does not tell us how an amp will sound, the simple fact is that in the time this paper was written until now measurement techniques have improved quite a lot, especially with the introduction of the Audio Precision analyzer. The real problem is people don't know what to look for. But if all the measurements are really there, its possible to predict the sound from what is shown in the measurements.
Sorry Ralph, but you are clearly not reading everything or overlooking important points.
1) This is not true, Cheever goes into quite some detail about the history of distortion measurement and the importance of higher order harmonics starting in the 1930s through modern attempts (Pages 7 - 20).
2) Not sure what point you are trying to make. Maybe your solution lowers distortion but I don't see dramatically lower distortion in most amps today than in the past...
3) I have read Bruno Putsey's papers but for me the proof is in the pudding. I detest the sound of his hypex and Mola Mola amps. So, his whole argument that you simply can't use enough feedback doesn't wash sonically with me. If it sounded really good then I might listen to his words. Until then...
4) Cheever is working on a metric to predict sound quality, not the specific details of the workings of some amps vs. others. If you use his metric correctly it would take that into account, or you can modify to take it into account very easily. BTW, there are TONS of modern amps that have rising distortion with increasing frequency...both SS and tube but primarily with feedback, whereas without feedback it is pretty flat with frequency in the upper ranges. Poorly designed tube amps (both PP and SET) will have rising bass distortion because of saturation effects. So, Cheever was not ignoring this (and it is important to the overall sound of an amp) but it was not relevant to his model.
https://www.stereophile.com/content/buckeye-purifi-eigentakt-1et9040ba-monoblock-power-amplifier-measurements
Notice the steep rise in distortion above 5Khz! That won't sound good and that is your Putseys in action.
Note also on the remeasurements the distortion is basically all odd order...something that NEVER correlates with good sound...no matter how low it is.
https://www.stereophile.com/content/burmester-218-power-amplifier-measurements
Note the rise above 2Khz.
The point is that all of this data will contribute to a worse score on Cheever's metric if you use it correctly (i.e. not only at one frequency)
5) No, because it also seemed cleaner as well as more dynamic. He further notes the listeners find the low powered triode to have less "grain", which is code for higher order distortion. The triode amp had less perceived grain, therefore less AUDIBLE high order distortion. If it was higher orders cause what you call a fake dynamic it would also sound significantly dirtier not cleaner. This is obvious when you listen to so called "low distortion" amps being pushed too hard and making a lot more high order distortion as they start to clip...the sound is not more dynamic but the exact opposite. It starts to sound congested and dirty and less compressed. It is you who is misinterpreting the perception not Cheever. BTW, he operated the amp within it's power bandwidth.
Given that higher order harmonics are perceived as harshness, grain and congestion, as well as reduced imaging space resolution, it is therefore illogical to attribute the SET amps superior perceived clarity, dynamics and spaciousness to an increase in high order harmonics.
6) We will have to agree to disagree. It is not a logical conclusion that you have come up with, sorry just isn't. You fail to realize that real instruments will have completely different profiles from the one the ear makes...like most amps do. Different makers of the same instruments also have clearly different emphasis on the overtones the instruments make, even though most of the harmonics are the same but which ones stand out more or less is different. I know this from making measurements on old violins myself. My ex had a Guarneri del Gesu, Stradivarius and Amati all at the same time in house. We measured them with a RTA and saw the changes as she played the same note with the same intensity. Just changing the bow also had a big effect on the distribution. So, even if the ear/brain filters some of that out while eliminating the self-generated harmonics there is still plenty left to distinguish it, not only from other instrument types but other makers of the same type of instrument.
The sheer unnaturalness of electronic distortions means that they stick out like a sore thumb. No amps are truly natural, except for a purely mechanical amplification like a grammophone. Electronics are inherently synthetic but the problem is exacerbated by circuit topologies that create harmonic patterns that are completely alien in nature. This the ear/brain just hears as artificial...regardless of how clean it appears on a scope.
I just got back from Munich where my wife attended all 4 days with me for the first time ever. I gave her the task of taking notes in each room with the goal of giving her a task to keep her engaged in listening. Do you know what the number 1 criticism was that she levelled against most rooms? She said "It sounds electronic and not natural." This applied primarily to rooms with low THD SS electronics...sad but true. She is not an audiophile but loves and attends regularly (with me) live classical, ballet and opera concerts.
The rooms she felt sounded the most natural had either SET, hybrid no feedback or PP tubes.
I detest the sound of his hypex and Mola Mola amps.
A lot of people use the Hypex module. I think its likely quite competent, but two things that can really affect the experience are the power supply and the input buffer. Those are often left to chance. If you skimp on a class D power supply thinking that its less important because of the efficiency, the amp will likely fall flat on its face. The design of the buffer can seal the amp to a similar fate. So you do have to be careful in making such assessment. I concede I've never heard a Hypex module so don't have direct experience in that regard.
FWIW distortion does not rise with frequency in our class D; its a ruler straight line across the audio band just like our OTLs. Both better than what a zero feedback SET can do FWIW...
If it was higher orders cause what you call a fake dynamic it would also sound significantly dirtier not cleaner. This is obvious when you listen to so called "low distortion" amps being pushed too hard and making a lot more high order distortion as they start to clip...the sound is not more dynamic but the exact opposite. It starts to sound congested and dirty and less compressed. It is you who is misinterpreting the perception not Cheever. BTW, he operated the amp within it's power bandwidth.
The phase of the third harmonic can cause superposition; increasing the sense of power of the fundamental. But if you think my position about higher ordered harmonics is false its very easy to prove with simple test equipment that it isn't. And I've also noticed that when you hear an SET being 'dynamic', if you also know its caused by distortion its much easier to hear that it really isn't all that clean. Knowing this can mess it up for you so sorry about that...
I don't think Cheever's metric is off base, just that it doesn't work the way he thinks it does. Occam's Razor sorts this one out easily: is the way SETs sound due to a really complex explanation about human hearing distortion as presented in his paper, or is it simply masking?
If you are a scientist, I'm sure by now you've seen scientists make mistakes. This happens because they are human. Of course proper vetting is supposed to make sure this sort of thing doesn't happen, but the process isn't fool proof.
I will only say that masking doesn't explain a lot of sonic observations, whereas Cheevers explanation makes a lot more sense. Others tried other metrics with only a masking concept, as highlighted in Cheever's intro, but these don't successfully match with listening observations. Occam's razor only applies when the simplest explanation actually explains everything...-in this case it doesn't.
Link below.
your argument is that a psychoacoustic effect caused by 2nd harmonics allowes one to perceive more detail and lower level detail???? Can't buy it. I have directly compared identically pure silver against copper TVC's, silver clearly lifts a vail and gives you more low level detail and more you are there, this is a passive device, something else is going on with a good tube amp
What really happens is masking occurs from lower harmonics being louder than the higher orders.
Lower ordered harmonics definitely add a coloration! Audiophile words like 'warmth' and 'bloom' describe this coloration, which is not objectionable unless it gets really profound. But don't fool yourself! The lower ordered harmonics are audible, and are not 'filtered out'. Wherever you got that is simply incorrect.
So its more accurate to say 'This means that what you are hearing, as long as you don't push the amp too hard, is effectively nearly pure because the lower ordered harmonics are not objectionable and mask the higher orders.'
Low order harmonics are audible only when they surpass the levels of self generated harmonics. Tests show 2nd harmonic is inaudible up to about 2%. for example.
If you are then the simple answer is he is simply incorrect.
You don't need me to hold your hand do you?
I saw this back in the day but I was less critical in my thinking back then so I will look afresh. Skimming some of it yesterday I was struck by how this is a Masters thesis from an EE department? Shouldn't psychologists be doing this work and, presumably hearing psychologists did not peer review it? None of the hearing data underpinning this work was created by him, and he goes from sweeping statement to sweeping statement. In my small perusal of this subject I found there were no sweeping statements but a lot of small steps and each one had some data that countered the main conclusion. He used the phrase 'all analog chain' at one point so we should be on guard as confirmation bias is a real thing in research. And when he talks of subjective results they are his and not unsighted panel results. This is more reviewer than researcher. Anyway, I shall read more and more thoroughly while staying aware of my own conformational bias that if the brain subtracts distortion from signals entering the ear how do we tell the difference between instruments and voices that get their character through distortion?
I can only say that having measured rare violins myself, the harmonic content of that particular instrument is very different from the monotonic decay the ear's mechanical mechanism makes. So, even if it did cut that out, there would be plenty to distinguish the rest...plus the levels are not really comparable. A violin playing at 85dB will be much louder than the ear's self generated harmonics.
However, I suspect that the wet supercomputer between your ears has other ways to determine the source and correlation between harmonics from the violin vs. harmonics from the ear mechanism. Could be related to phase adn locality proximity etc. etc. This could also then be why such small levels of amplifier distortion are still having an audible impact.
Maybe. Occam's razor might suggest that the addition of distortion provides the character that is missing from low distortion designs. However, I don't think the simplest explanation works in the audiophile world, it is usually the most expensive solution :)
I don't agree that this is the simplest explanation. To agree to this explanation is to then assume that nearly all recordings are deficient with regard to realistic sound quality. Is that likely? I think not.
A simple explanation following Occam's razor should not then open up a can of worms around the recordings.
For me the simplest explanation is that the most correct sounding amp must therefore be the one that is most free of audible distortions. The objective numbers are not so relevant.
IMO, believability in audio from the electronics is the absence of audible artifacts not necessarily measurable ones.
It also becomes clear to me that when a system makes most recordings sound grey and lifeless then it is not the recordings but the gear reproducing them...something is psychoacoustically off.
Whether you disagree with the idea of SETs adding something, as you do, I don't think it strengthens your argument if you fail to see that as a simpler explanation. Audiophilia does tend towards the exotic and expensive and, perhaps, that extends to theories also. I noticed that you did not simply measure the harmonic content of real instruments but of 'rare violins'.
Anyway, back to the topic in hand. It seems Cheever took his aural harmonic generation creation data from Olsen. A copy of that book should arrive today...
Read the Keith Howard experiment in Stereophile where he added different distortions to audio files. He found that NO added distortion sounded best but the least worst was alternating even and odd in an exponential decay. The worst was all odd harmonics, like most PP amps produce.
You used to be able to download the code and play with it yourself. I did and came to essentially the same conclusion as Mr. Howard.
I enjoy all of Keith Howard's writings. IIRC, there is not a big difference between any of the patterns and he struggled to order them.
There is a utility that allows you to add a desired amount of distortion see link. If I want to add a lot of low order distortion I simply play a record.As to Cheever, I obtained Olsen's text book referenced by DC and it contains the same graph but Olsen attributes it to someone else. I obtained that paper and though relevant it is not the source for that graph so more searching required. What I can say is that Cheever misinterprets that graph - it shows the level of a fundamental at which it's harmonics are detectable. They are not the absolute level of the harmonics as the author suggests! The next graph is probably the most important, similar data attributed to Olsen but that must come from a different book. I will seek it out to find how the testing was done.
Edits: 06/01/25 06/02/25
I think he didn't think it through. I explained why elsewhere on this thread.
I have no additional technical input to offer about harmonic distortion, soft clipping, etc.Could microphonics in tubes be a contributing factor in hearing things that you don't normally hear in other amps. I often wonder myself if this contributes to that so-called 3 dimensional sound that tube amps are so good at creating.
Edits: 05/18/25
for me has been damping rings on small signal tubes. Many preamps available since the late 80s sport such.
I find they assist the VTL amps with the NOS Sylvania AT7 and RCA 6350 tubes on some content to minimize tizz on top. Amps definitely get a piece of the back wave. As I find with better power supplies, a bit of fog is removed for improved transparency.
Not denying that dampers help minimize microphonics. I've had them in some of my own tube gear.
My inquiry and curiosity is whether microphonics contribute to the 3 dimensional tube sound that some crave.
I used Herbie Halos when I used tube equipment.
Regards,
Steve
Factory dampers on ARC pre.
Could be. These were popular.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
.
Actually there is a measurable reason for the high 2nd harmonic distortion(a truly lovely distortion for real) of SET amps. They are non-linear. They amplify asymmetrically. If you put in a simple, single frequency sine wave you can see this on an oscilloscope. The positive side of the sine wave goes higher than the lower side goes down. And this is true for all wave forms including music. This is inherent in the topology of single ended and can't be designed out therefore.. It's true for solid state single ended amps also.
The only reason SE amps have even harmonic distortion components is because the even harmonics are not cancelled out by the circuit topology. All other amps are differential and the even harmonics are naturally cancelled out by again the circuit topology through the load. The sinusoidal waveforms sent out from instruments have both even and odd harmonics inherently.
and the power triodes typically used in SET amplifiers are supremely linear devices.
SE transistor amps'll do the same thing as an SET (in terms of spectrum of harmonic distortion products). Cf. Nelson Pass's "Amp Camp Amp" -- decidedly SE, and decidedly solid state. :)
(an Amp Camp Amp configured as an actively cooled, stereo amplifier powered by dual Dell power bricks & based on the 'Bottlehead' amp kit family's form factor)
all the best,
mrh
You are correct. Triode output tubes are extremely linear devices. It's circuit topology that is responsible for the dichotomy between the positive and negative halves of the wave form.
Congrats on a nice Nelson Pass build of a solid-state SE amp. I have an original First Watt F1 amp signed by Nelson himself. Solid state devices are not as linear as tubes in general, so a feedback circuit is required, even in your amp, for improved linearity. Not so, with a 300B SET. There is no fb at all. Accuracy is totally dependent on the linear characteristic of the tube. Other tubes used in SET designs are (in increasing wattage capability) 45, 2A3, 300B, and 845. These are of course inherently Class A. I have mono 300Bs driving Cain & Cain Abbeys.
The biggest difference is that you can add gain more easily in a solid state amp so global feedback can be more effective.
"The biggest difference is that you can add gain more easily in a solid state amp so global feedback can be more effective."
In my opinion at that point you have thrown out the baby with the bath water.
Unless it is pure feedback (with nothing reactive in the loop) it reduces the overall percentage of the HD but increases the order and the higher the order the less of it it takes to sound bad.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
I think a fundamental question an amplifier designer has to ask is 'what am I trying to achieve'. Minimizing distortion is a valid approach, IMHO, and if so, that is the bath water the baby will enter. Does that approach sound better or does it rob the sound of some 'flavor' even if it is more accurate because some color is preferred? Low feedback circuits produce high harmonics too but they usually have much more low harmonic content and are they preferred because the low harmonics mask the high harmonics or does the larger harmonic content simply provide that flavor?
I firmly believe that people should choose what they like. If you don't want to bathe the baby at all that is just as valid and, probably, easier to achieve. The trouble comes when people can't just enjoy what they like for its own sake but must pronounce it objectively superior for some reason.
But amplification is not totally defined by low IM and harmonic distortion which are the main things we measure. There are other factors involved in good sound and if they aren't good the low distortion is a wasted element. Why do you think Gordon Holt founded Stereophile and Harry Pearson founded TAS? Before that amp testing was frequency response and distortion. And good tests often didn't go along with good sound.
We know (or at least I do as I read Stereophile for a few years) that measurements and subjective response may not correlate. And that is fine if a bit perplexing and my interest is that gap. The guy who said 'if something measures well and sounds poor you are measuring the wrong thing' deserved a knee in the groin for not finding out what those things are. You suggest that you know so please tell.
I've never read TAS and Gordon Holt pre-dated my Stereophile readership (my money was on the erudite Englishman). Perhaps they started their magazines to make a living listening to records? I'd take that job though if you started to hate your work you'd have nothing to escape it (perhaps do a little accounting on the side?). Distortion testing tells you how much fidelity the output has to the input so it does seem an odd thing to say that 'high fidelity' can be wrong? Is there some quintessential element that is immeasurable that is passed on only by select components or do they add something pleasant? It would be nice to know the answer. Maybe some people do but keep it a trade secret and, if so, maybe it should remain a secret otherwise good sound will become commoditized and there would be nothing to post about.
The guy who said 'if something measures well and sounds poor you are measuring the wrong thing' deserved a knee in the groin for not finding out what those things are.
That was the head engineer of HH Scott, Daniel vonRecklinghausen. Not sure if he didn't find out; the HH Scott stuff has a pretty good following.
I'm of the opinion that if you know the harmonic spectrum at 1 Watt and 6dB below full power, if you know the distortion vs frequency then you can get a pretty good idea of how an amp will sound.
BTW you mentioned something about lower ordered harmonics masking higher orders. This sort of thing has been known in speakers for a long time; I see no reason why the ears's masking of harmonics with a speaker or an amp would be any different.
"I'm of the opinion that if you know the harmonic spectrum at 1 Watt and 6dB below full power, if you know the distortion vs frequency then you can get a pretty good idea of how an amp will sound."
That would seem to neglect the performance as power goes to 0, which is what defines the all-important sonics of decay trails, nuance, etc.
Amps that do not exhibit a linear decrease of distortion into the noise floor are to be avoided.
(And yes this is also common knowledge.)
I should have mentioned 1 Watt performance too.
Non-linearities create distortion such that the harmonics change differently in level compared to changes in the fundamental. Second harmonic level changes, if the fundamental level is proportional to x the second harmonic level changes as x^2, the third as x^3 etc. I've wondered if natural sounds do the same, so do the relative levels of the harmonics of instruments change in the same way? I don't know if that is the case but if it is true I wonder if the best thing for an amplifier to do is to mimic that behavior. Some power amps measured by Stereophile show a straight line of THD vs power (on a log-log scale) with a slope that shows which harmonic is dominant, and sometimes the slope changes as power increases. However, most don't do that. Maybe that would be a good indicator of 'natural sound' to have the harmonics decrease in a non-linear fashion as the signal falls to zero.
(Note, this is the concept of distortion intercept points, a metric that is used in the RF world to predict distortion based on one measurement taken at a low level knowing it will follow a predictable slop with input power)
This is basically the hypothesis Cheever has proposed. Amps generating distortion that mimics the pattern of the ears own distortion mechanism will "hide" in that self-generated distortion...ie. it's masked.
Inspired by some conversations here, last year I read a book on hearing psychology. I was particularly interested about masking and it was covered but not specifically about how low harmonics might mask high order ones. It would be really interesting to quantify the effectHere's a thought experiment: a 1kHz signal generates 1% second harmonic as well as all the harmonics above that and we suppose that the second harmonic is large enough in amplitude to mask the fifth harmonic. Whatever that amplitude difference, they are separated in frequency by 3kHz. But, the fundamental is only 4kHz away and it is 40dB larger than the second harmonic. It seems more likely that the fundamental would mask more?
Edits: 05/23/25
As you probably expect, Cheever's explanation is quite complex compared to that of simple masking. Occam's Razor says the rest.
To answer your question, 'it depends'. The amplitude of the 5th compared to the 2nd or 3rd will tell whether its masked or not. The same is true of the fundamental compared the the 5th.
Our ears use the higher ordered harmonics to sense sound pressure. So its very keen to their presence. On this account the further from the fundamental, the more important it becomes to be masked as the the higher the order, the more annoying it is.
Not everything in science is Ovcam's razor...-you overuse this and I am not surprised because most non-scientists don't realize the simplest explanation is actually not correct...-otherwise scientists wouldn't have much to do and we could leave it all to armchair scientists who think it's all so obvious...-when in fact it's not.
So far this hasn't been about science so much as simply trying to make someone wrong.
When that happens Occam's Razor is pretty useful.
Sorry about the title, couldn't resist. Not sure what you meant about high harmonics indicating sound pressure.
I do think discussing these masking qualitatively is a bit of a get-out. Maybe the best place to find more quantitative results would be the mp3 group as they calculate what is masked and what can be discarded. High bit rate aac is pretty good (IMHO) so they must be in right ball park.
I re-read some of the one hearing psychology book I have and in the section on hi-fi specs the author comments that low order harmonics are likely masked by the fundamental but the higher ones are not masked at all. I find that more credible than, say, 2HD making high order harmonics. Wish I had some worked examples.
to sense sound pressure.
You can demonstrate this for yourself with simple test equipment. You need an amplifier, speaker, some sort of meter (VU meter works best, you can also use a sound pressure level app on your phone) and a sine/squarewave generator.
Run the sine wave thru the amp and speaker and set to a comfortable level. If using a VU meter, 0VU.
Cover up the meter and turn down the volume, switch to square wave and then turn up the volume until it sounds just as loud. Uncover the meter and you'll see that the ear uses higher ordered harmonics to sense sound pressure.
The fact that a 3rd harmonic can mask higher orders in loudspeakers has been known for some time. Its no different in amplifiers. The key is the relationship between the various harmonics. The 2nd or 3rd must be prodigious enough compared to the higher orders that the ear only hears the 2nd or 3rd (as a 'warmth' coloration, which is not annoying).
3rd can only mask the fourth...-fourth the fifth...-.fifth the sixth etc...- but ONLY if the preceding harmonic is significantly higher in amplitude than the following harmonic.
If an attorney were litigating this he'd be asking you if you were lying then or lying now?
Put another way you can't have it both ways. We both agree that the harmonics should fall off in amplitude with the order of the harmonic according to an exponential function.
But in your other post regarding Cheever you take a different argument which contradicts this one.
I find when people do this sort of thing its not about actual fact as it is about making someone else 'wrong'.
No, I don't see what Cheever states as a contradiction...perhaps you just misunderstand how he is using Aural harmonics?
-
I now understand your point but, don't forget, that a sine wave and square wave have different peak-to-average ratios so it is not surprising they are perceived differently.
Back to the pedantry, the ear's sensitivity to sound pressure and our perception of loudness are different things. The pressure for 'just detectability' gives us the well known Fletcher-Munson curves where the y-axis is in pressure but hearing psychologists measure loudness in phons where test subjects compare subjective loudness of different sounds, like your comparison of sine and square waves. But to get to that point the sound pressure has already been 'measured' by the ear and now it is about signal processing.
As to masking, again, I am just curious to see some numbers for my own education. I would be grateful for any references that quantify the effect
You might try that test anyway since an analog VU meter simply uses the energy available. If you don't care to do so, FWIW the VU meter tends to read about -25dB which is not explained by the difference in square and sine energy.
Any musician that works with waveforms or distortion knows how that can affect the instrument's ability to cut through the mix.
Fletcher Munson affects this since if higher ordered harmonics show up where the ear is most sensitive that doesn't help with how the amp might 'sound' since they are likely to be more audible.
For basics on masking see the link.
Cheever states on page 40 of his thesis: "The aural harmonics play a more prominent role than the ears masking mechanism. There was no correlation between age and sex in these or other studies of aural harmonics." He defines aural harmonics on the preceding few pages before page 40.
He also notes that the aural harmonics order and amplitude are dependent on SPL. So, SPL needs to be accounted for in the metric because the ear gets less sensitive at higher SPL because of the increased amplitude and order of the aural harmonics.
It is interesting to note that Cheever makes a prediction (remember this thesis is from 2001) that single ended designs will continue to become more and more widespread. He was bang on because I saw a whole lot of SET tube love at Munich this (and last few) year.
Jean Hiraga hit upon it a long time ago. Cheever expanded this argument with more technical backing. Cheever comes up with a metric that takes SPL into account because your ear's own harmonics change with SPL.
Basically your ear produces 2nd harmonic distortion at only 15dB below the main signal at 80dB and only 12 dB down at 90dB. 3rd harmonic is only 35dB below an 80dB signal.
This allows for distortion to be masked by the brain's filtering of the ear generated harmonics.
This is basically the only way that SET can sound not only sound pleasant but actually sound very pure.
I am a scientist and for a long time I thought that high power and low distortion must be the right way to go...except for my pesky ears telling me that this wasn't getting closer to what live, unamplified, instruments in space sound like...and there was no convergence with better measurements.
Then I heard amps that didn't rely on ultralow distortion but instead used tubes and then SETs with no feedback. Once I banished feedback from my system and went all single ended things got much more interesting and sonically recognizable with the real thing.
So, my experiments with my ears trumped what I "knew" and then I set off to find out how it could be possible. This is where I started to realize that the ear/brain behaves completely different from measurement equipment.
Realizing this was a revelation that the best measurements for hearing humans is not the same as for an oscilloscope (or other dedicated measurement system). How else could a model T Ford, I mean SET, keep up with modern equivalents?
This was again reinforced this past weekend at the Munich High End show where hands down the best sounding rooms either had SET, TriodeFET (single stage single ended), Class A PP triode or Class A hybrid with no feedback...and usually, but not exclusively, horns.
The 90 year old WE speakers were coupled with a 1 watt amp from Silbatone from Korea (the man has a WE museum of his collection in Seoul). Despite the obvious horn colorations from this antique system, the sound was incredible and truly lifelike. Maria Callas was in the room!
The worst sounding systems were the ones with low distortion SS amps...sorry they simply sound, as my wife said (she attended all 4 days!), "electronic" as opposed to "natural".
Great post!
Something else not discussed in this thread is this: We actually don't know how amps behave with music. We measure them with sine waves (maybe a couple at a time for IMD). We are largely making an *assumption* that an amp's distortion spectra with a highly complex signal (music) is identical to that with sine waves.
We actually don't know how amps behave with music. We measure them with sine waves (maybe a couple at a time for IMD). We are largely making an *assumption* that an amp's distortion spectra with a highly complex signal (music) is identical to that with sine waves.
If the amp has improperly applied feedback, I've long theorized that it could behave chaotically with a strange attracter around its phase margin. Non-linearities in the speaker contributing to the behavior, depending on how well the feedback loop is designed.
Zero feedback amps don't have this problem but you also don't have a really good idea about how they are distorting either, although we sometimes have a pretty good guess based simply on the sine wave response.
The 32 note test signal is a better means to see what is really happening.
Here is a 32 tone test signal, watch from 14:08 to 15:48. A signal like this can have a peak to average ratio higher than music. He doesn't show what is looks like in the time domain but it will look nothing like a sine wave and more like a music signal
A more valid approach is to take what has been learned about psychoacoustics and apply that to make amps with distortion patterns that can "hide" in the ear/brains own masking system. SETs are already going in that direction but with some clever application I bet it would be possible to make amps that mimic almost perfectly and then will sound truly invisible "straight wire with gain" paradigm.
Called mp3 :)
It's called SETs ;-)...-
because about 99% of all speakers made are 'Voltage driven'.
A 'straight wire with gain' is a poor analogy and really should never be used since it can never ever describe an actual amplifier.
Tube amps without feedback try to act more like a power source than a Voltage source. That is why with most speakers their frequency response isn't flat. That is also why before the Voltage rules were adopted in the 1950s, most speakers had controls on the back to allow the user to adjust the speaker to match the power response of the amp since its output impedance was unknown and high (by today's standards).
Is it really true that most speakers prefer a voltage source, that halves power into higher impedance?
I just read an (ancient) review of the Tom Evans Linear A SE amp (a rather unique affair with op-amp drivers and parallel, quad EL84 output) and Tom asserts that the fact that his TX-coupled amp produces constant power across a relatively wide impedance range will make it sound more natural on the typical (dynamic) speaker.
Transformer-coupled tube amps' high output impedance is usually given as a *negative,* but it seems it's about the opposite. I guess this is all about voltage vs. power - a transistor amp will *lose power when speaker impedance increases* - but that speaker impedance has nothing to do with the input signal.
Keeping the power roughly the same when the speaker impedance varies seems much smarter. Is my reasoning off somewhere?
Thanks.
What are examples of speakers that don't follow this paradigm? Electrostats?
certain horn speakers- usually identified due to crossover level controls on the rear, meant to allow the user to tailor the speaker to the power response of the amplifier, which will be different according to the amp's output impedance.
Is it really true that most speakers prefer a voltage source, that halves power into higher impedance?
Yes.
Keeping the power roughly the same when the speaker impedance varies seems much smarter. Is my reasoning off somewhere?
Whether its smarter or not isn't the issue. Back in the 1950s EV and MacIntosh developed the Voltage drive rules since this allowed greater plug and play. If you wanted constant power with respect to impedance, you either ran a zero feedback tube amp, or one that had equal amounts of Voltage and Current feedback. There were tube amps made with 'damping' controls which allowed you to do exactly that, or have all current feedback (very high output impedance, often a multiple of that of the speaker) or all Voltage feedback (very low output impedance).
IOW the Voltage rules are what guides most speaker designs today. Tom Evans was simply incorrect in his statement; if the amp really was constant power he would have tonal anomalies on most modern loudspeakers.
N.T
-
Just so I'm clear, are you saying that all solid state amps use feedback? As an F1 owner, I'm sure you're aware that mister Pass says it uses no feedback. (see his quote below) Whether that's truly unique to this SS amp only, I don't know.
What I do know is that my First Watt F1 (since updated to F1J by First Watt) is the best sounding amp that I own. I kinda felt like a traitor to tubes at first, but I quickly got over that. Note that I've always had push-pull tube amps though, never a SET.
"If being a current source amplifier isn't different enough, the F1 is special in other ways. It uses no feedback to reduce distortion, flatten frequency response or create a low output impedance, and it ignores the voltages that appear across the speaker terminals." Nelson Pass, from his F1 power amplifier paper.
Some solid state amps go without feedback; HOWEVER, solid state devices (MOSFET, JFET, VFET, SIT, BJT etc.) are inherently less linear than triodes. this is a fact of physics.
Supposedly, the SIT are closer to triodes than other transistors and MOSFETS behave quite differently from BJTs but none of them have the same transfer function as a triode.
This means even single ended and without feedback, they WILL NOT sound the same as a triode amp. They will have a somewhat different character that will include less desirable harmonics emphasized vs. triodes.
Actually the curves on the old Sony VFETS look a lot like the curves on a good triode.
Do I have read as well. However, everyone I know who had tried them in the FirstWatt configurations say they sound KIND OF like triodes but not really. Is it just because of somewhat lower overall distortion? Is it because of transformer coupling?
I had a single ended hybrid from NAT that used a single large MOSFET for the output stage...-no feedback and no output transformer. It sounded crazy good in some ways but not as natural as some of the best SETs I have owned.
Now I use a different kind of hybrid that is not only single ended it is even single stage and combines a triode and FET into a single amplification element. Transformer coupled and no feedback . Sounds like a very clean SET.
-
nt
-
Well, to be sure the F1 is an extremely unique amp. It uses an impossible to find FET (actually two of them) which tells me this is not SE, in fact Nelson Pass calls it balanced single ended. It doesn't need feedback because it forces the current in the output to follow the current reference, but it also demands a single efficient speaker magnet (in other words a full range single coil speaker). So that's pretty special. I use my F1 to drive a single full range Fostex (Cain & Cain Abbeys). But I do have 300B SET monoblocks also and I like the sound with them better than the F1.
I think all Pass amps do not employ full loop negative feedback including his production amps.
He uses some in current X series amps to provide the "best sound".
"With the X250 we applied GNFB and have used it in the X amplifiers since."
By the way it's Nelson Pass who confirmed to me the dichotomy between the positive and negative halves of the wave form.
The ACA's a toy (and, in full disclosure, I didn't build it; a fellow on the Polk forums did) - not a bad one, though. I mentioned the linearity of power triodes, and if you'll look again at my post, you'll note that I qualified the behavior of SE transistor circuits as similar to SE tubes " in terms of spectrum of harmonic distortion products ". ;) Yes, (most) transistors aren't nearly as nicely behaved as a triode in terms of linearity.
I actually didn't build my daily driver SE 2A3 amp (Joseph Esmilla's design, slightly tweaked), either -- but the builder and the designer are both well known by me (and vice versa ). ;)
I have built a few things -- they tend towards the ugly but functional, though.
all the best,
mrh
Try a 2A3 amp with interstage coupling...I was shocked at how much more drive this gives and made the amp sound huge compared to it's rated power of 3.5 watts (I have a Silvercore "Silberglut" 2A3 that uses 6C45PI tubes interstage coupled to the 2A3s).
Are you asserting that this explains the extra nuance I'm hearing, which seems entirely natural?
I have no idea how audible what you describe is, for one thing.
(But thanks for this. I was not aware of this technical fact.)
| FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
||||||||||||||
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
