Mastersound PF100 PSET Amplifier
- Thread starter Ron Resnick
- Start date
-
- Tags
- mastersound pf100
Calling someone a religious zealot in a clearly derogatory context is not an observation…it’s an attack. You have no idea my audio history and the number and type of amps I have gone through over the years. Your lack of context makes the attack even more egregious. I still really like the sound of some hybrid amps and some push pull triode amps and OTLs. The new compound element amps (like Aries Cerat and NAT produce) could take things beyond SET. I am an experimental scientist and based on my observations and study into the subject this is what I find. You can disagree but you can’t say what I think is based on blind faith…it most certainly is not.
If you are talking about amps of the same design then perhaps talking about 1-2% distortion audibility might have some merit but once you start mixing topologies all bets are off.
I described two methods. One in which the only variable is an OPT tap point in order to alter winding ratio and thus output tube load line, the other an inline THD generator which allows you to inject and dial up/down THD in a given system chain. In both cases, I'm listening to (and measuring) THD changes within an otherwise static chain. The first method works well until or unless you run out of power, which never happened with 98 dB/mW headphones. None of this was done in order to prove a point to anyone, rather just to educate myself as I'm currently working on the design and build of DHT-based component for my system.
Are you talking about the PVW Stereophile article or Geddes? Geddes has both slides and actual publications
I state that -6dB point, giving a lot of amps the benefit of the doubt. My experience mirrors yours in this regard- many SETs have less usable power than 20%!
Much of what you state here is false. IMD is a very audible distortion subjectively (and has a lot in common with the way the ear perceives aliasing) and this fact is well-known and should not be a topic of debate! The ear uses higher ordered harmonics to sense how loud sounds actually are, and the ear is a range in excess of 120dB. So it is keenly sensitive to the higher orders as well. This is very easy to demonstrate with simple test equipment! If you look at musical instrument waveforms (with any accuracy), you'll see the higher orders play a role in how those instruments sound. The idea that distortion isn't audible is really ridiculous: harmonics define the sound of any instrument and by distorting the harmonic makeup of them, you change their sound. So often harmonic distortion is treated as something that is so many 'dB down' from the actual signal- I find this really amusing since the signal itself is being modified by that distortion.
The difference between a solid state amp with higher ordered harmonics and an SET is the solid state amp maintains low distortion as power is increased. SETs do not- the distortion rises fairly linearly (and relatively speaking, dramatically) as the power is increased. Feedback is preventing this behavior in solid state amps. Their brightness and harshness comes from two things. The feedback signal being distorted before it can even do its job due to it being applied to a non-linear point in the amplifier (usually the input of a semiconductor) and the circuit usually lacks the Gain Bandwidth Product to support the feedback over the entire audio range- so distortion is higher at high frequencies- usually right in the range of the ear's most sensitive area (Fletcher Munson).
So this behavior of SETs is very real and easy to measure. Its also easy to hear- and once you know that 'dynamic' quality is coming from distortion rather than the signal (which is where it should come from) it kinda wreaks it. Again: the ear perceives more higher ordered harmonic distortion as increased sound pressure which easy to demonstrate with simple test equipment (procedure available on request).
This is utter BS. Nonsense. Either a measurement error or he didn't realize that distortion makes tube amps seem to have more power than solid state due to how our ears interact with distortion (as I've written about ad nauseum). SETs, lacking feedback, can behave differently with regards to voltage vs load since they behave more like power sources than they do voltage sources. So when working with them you will see the voltage rise with the impedance of the load- but the truth of the matter is revealed when you calculate the actual output power. So it can appear to behave differently than solid state, which tends to act as a voltage source. More on this phenomena at http://www.atma-sphere.com/en/resources-paradigms-in-amplifier-design.html
Ears are different and people have different taste, but the rules our ears use are the same across the entire world population. For example, our ears use a logarithmic scale for sound pressure rather than linear- which is why we use deciBels to denote sound pressure. How we detect sound pressure is the same. Our sensitivity to higher frequencies is the same (hence musical instruments are built the way they are). If our ears used different perceptual rules it would be impossible to design audio equipment plain and simple.
One must be careful to not conflate the rules of human hearing perception with taste!
Stop referencing yourself in these debates!
Go read the Geddes paper, they found no more correlation with IMD than THD. Of course the harmonics matter!! On this at least we agree.
I did.
There are some problems. For example, THD is something very specific while IMD is not. Exactly what the procedure was is not documented in the slides which are meant to go with a talk. I've already pointed out that THD easily hides dirt under the carpet and explained the mechanism. IMD OTOH simply is short hand for 'InterModulation Distortion' so lacks the specific nature that THD does not. Depending on how they set up the test the IMD might have been harder to detect- for example what tones were used, the nature of the 'non-linearity' involved and so on.
Its not as cut and dried as made out. But we have to look at Occam's Razor when considering this issue; IMD has long been recognized by both the objective and subjective parties as being audible and irritating- that is one of the few things on which these parties agree! This has been the case since the 1950s. In face of that with one study that might seem to show otherwise Occam's Razor suggests that one study might simply be wrong- that there was an error of some sort. Or- just as likely- a wrong conclusion was reached by someone.
Geddes has said elsewhere the perception of IMD increases with sound pressure. That would seem to fly in the face of his slide show so I think the 'wrong conclusion by someone' is correct. In this case I don't think that someone was Geddes.
Sorry Ralph, you didn’t read the PAPER. It states clear IMD is based on a combination of 100 and 6khz sinusoidal waveforms.
He makes clear that some distortion types are clearly audible at 0.01% but others at 10% are inaudible, rendering THD pointless.
The slides are just a summary of the papers
I agree that THD is pointless and responsible for lots of confusion- amps that sound good but measure poorly and the opposite.
Thanks for the links. A lot of what he says about THD, masking and the like is stuff I"ve been saying for years. So if I reference myself I've got good company
His comments about IMD IMO are incomplete. Its not enough to use the tones he did- real world music tends to have far more.
Personally I've found that when you reduce IMD in the circuit the presentation becomes less 'busy' (seemingly more 'relaxed') and its easier to hear into the sound stage. It varies depending on what is inter-modulating- the nature of the non-linearity causing it!!
So that's a tricky one to nail down. In some cases the IMD many well be in audible and in other cases it will affect the presentation. I've no doubt this is why some amps seem to be OK when the music isn't complex but fall apart when it is.
IMD has the same harmonic weighting problems as THD...which is why a simple metric doesn't work. An amp can have low order IMD components of relatively high level and sound less "busy" than one with relatively low amounts of high order IMD products.I agree that THD is pointless and responsible for lots of confusion- amps that sound good but measure poorly and the opposite.
Thanks for the links. A lot of what he says about THD, masking and the like is stuff I"ve been saying for years. So if I reference myself I've got good company
His comments about IMD IMO are incomplete. Its not enough to use the tones he did- real world music tends to have far more.
Personally I've found that when you reduce IMD in the circuit the presentation becomes less 'busy' (seemingly more 'relaxed') and its easier to hear into the sound stage. It varies depending on what is inter-modulating- the nature of the non-linearity causing it!!
So that's a tricky one to nail down. In some cases the IMD many well be in audible and in other cases it will affect the presentation. I've no doubt this is why some amps seem to be OK when the music isn't complex but fall apart when it is.
A NEW METHODOLOGY FOR AUDIO FREQUENCY POWER AMPLIFIER TESTING (next-tube.com)
The first section does a good job of outlining the problem and the history of attempts to understand objective vs. subjective data. Whether you agree with his approach to the development of a metric or not, his outcome is nodit very dissimilar to Geddes but is more usable and I think importantly takes SPL into account. If you can generate a high SPL from low power on an amp that performs best at low power then it is likely that most of its sins will be masked and therefore earns a good score in Cheever's metric. Put that same amp on a low sensitivity speaker and you will A) generate more higher order harmonics as the amp is pushed and B) not have the added benefit of masking that expands with SPL as the ears own distortion is increasing and creating greater "cover" for the electronics generated distortion.
What I have discovered though is that even people who listen pretty loud do not listen with average levels above 85dB with peaks maybe around 100dB. A reasonably sensitive pair of speakers (say 90dB) actually gets to these volumes on a few watts. So a 90dB speaker in most people's room will get loud enough without excessive power draw and a 20 watt amp is likely sufficient for 99% of their listening...and that amp cruising at a watt or less most of the time.
Look at Figure 2.3. At 90dB, your ear is producing nearly 10% 2nd harmonic and only 0.1% 3rd harmonic. This is part of the reason why even 3rd harmonic is not masked nor masking other frequencies very much. However, we see that self-generated 2nd harmonic is essentially invisible to the ear brain probably well above 1%.
Figure 2.7 shows his metric's output. According to his metric, at 70dBA, 1% of 2nd harmonic is invisible to the ear/brain...with normal speakers this is mW of power from the amp and nearly all SETs and other zero feedback amps will have far <1% 2nd order at this SPL level. Same for 80dBA, except now it is more like 2-3% 2nd order that is masked. But what about 3rd order? Well, it looks like at 70dbA the distortion needs to be at 0.01% or it is not following the pattern and will be exposed. 4th, 5th etc. are 10x progressively lower with each harmonic order increase. You were always arguing that 3rd masks like 2nd but this data clearly indicates that 3rd order does not effectively mask like 2nd order and the tolerability is approximately 100x lower for 3rd than 2nd order. Without the dominant 2nd and strict adherence to this distortion profile, an amps distortion will not be effectively masked and lead to audible differences.
The first section does a good job of outlining the problem and the history of attempts to understand objective vs. subjective data. Whether you agree with his approach to the development of a metric or not, his outcome is nodit very dissimilar to Geddes but is more usable and I think importantly takes SPL into account. If you can generate a high SPL from low power on an amp that performs best at low power then it is likely that most of its sins will be masked and therefore earns a good score in Cheever's metric. Put that same amp on a low sensitivity speaker and you will A) generate more higher order harmonics as the amp is pushed and B) not have the added benefit of masking that expands with SPL as the ears own distortion is increasing and creating greater "cover" for the electronics generated distortion.
What I have discovered though is that even people who listen pretty loud do not listen with average levels above 85dB with peaks maybe around 100dB. A reasonably sensitive pair of speakers (say 90dB) actually gets to these volumes on a few watts. So a 90dB speaker in most people's room will get loud enough without excessive power draw and a 20 watt amp is likely sufficient for 99% of their listening...and that amp cruising at a watt or less most of the time.
Look at Figure 2.3. At 90dB, your ear is producing nearly 10% 2nd harmonic and only 0.1% 3rd harmonic. This is part of the reason why even 3rd harmonic is not masked nor masking other frequencies very much. However, we see that self-generated 2nd harmonic is essentially invisible to the ear brain probably well above 1%.
Figure 2.7 shows his metric's output. According to his metric, at 70dBA, 1% of 2nd harmonic is invisible to the ear/brain...with normal speakers this is mW of power from the amp and nearly all SETs and other zero feedback amps will have far <1% 2nd order at this SPL level. Same for 80dBA, except now it is more like 2-3% 2nd order that is masked. But what about 3rd order? Well, it looks like at 70dbA the distortion needs to be at 0.01% or it is not following the pattern and will be exposed. 4th, 5th etc. are 10x progressively lower with each harmonic order increase. You were always arguing that 3rd masks like 2nd but this data clearly indicates that 3rd order does not effectively mask like 2nd order and the tolerability is approximately 100x lower for 3rd than 2nd order. Without the dominant 2nd and strict adherence to this distortion profile, an amps distortion will not be effectively masked and lead to audible differences.
Last edited:
I read that particular paper several years ago.
What I have quoted above is exactly the problem we see with most people's implementation of SETs and is what I've maintained all along. Nice to see your (or Cheever's) agreement here. I essentially said the same thing as the text above in this post: https://www.whatsbestforum.com/threads/mastersound-pf100-pset-amplifier.24351/post-891230
I don't agree that THD is pointless. But it absolutely can be utilized incorrectly to (try to) compare apples vs tangelos and cause confusion. It can be a useful measurement to utilize in the voicing of a given topology and tube family, primarily because it's easy to measure and changes in THD are often easily audible. So once you know where you want your design point, then you can work towards it through simple measurements - which is convenient.I agree that THD is pointless and responsible for lots of confusion- amps that sound good but measure poorly and the opposite.
Thanks for the links. A lot of what he says about THD, masking and the like is stuff I"ve been saying for years. So if I reference myself I've got good company
His comments about IMD IMO are incomplete. Its not enough to use the tones he did- real world music tends to have far more.
Personally I've found that when you reduce IMD in the circuit the presentation becomes less 'busy' (seemingly more 'relaxed') and its easier to hear into the sound stage. It varies depending on what is inter-modulating- the nature of the non-linearity causing it!!
So that's a tricky one to nail down. In some cases the IMD many well be in audible and in other cases it will affect the presentation. I've no doubt this is why some amps seem to be OK when the music isn't complex but fall apart when it is.
If IMD is pointless too, then we shouldn't worry about painstakingly aligning our cartridges to minimize tracking error because it shouldn't matter as it primarily manifests as IMD. Strangely, though, minimizing tracking error always seems to correlate with subjectively better sound to my ears.
Yeah, I know that "it all depends". The point being, we're starting to paint with too broad of a brush on this topic, in this thread.
Cheever even makes the point in his intro that when all amps were no feedback (it hadn’t been invented yet) and basically SET (with some notable exceptions) then THD had some meaning.
Once feedback was introduced…all correlation was lost.
Regardless of how the amp is designed, according to Cheever it has to have a harmonic pattern that follows the aural harmonics or it will sound distorted in some way…unless it has truly zero distortion. What he observes, and I agree by surveying measurements from Stereophile and Sounstage, is that NO modern amp designs have that distortion pattern…only SETs approximate it. Jean Hiraga came to a similar conclusion about the distortion pattern.
For DHT SET, I generally dislike the sound of feedback. I would probably take higher distortion and no feedback over lower distortion with feedback most of the time. Assuming damping factor isn't critical. Best of all, I've found that I prefer to limit usable power so as to obtain low distortion with no feedback. This of course has its drawbacks, such as practicality, but has its place in certain line stages, headphone amps to name a couple.
What he did not take into account is two things:
1) how the feedback is applied; in most cases its applied to the cathode of the input tube, which will result in the feedback signal being distorted. This results in higher ordered harmonic generation, causing amps with feedback used in this manner to be harsher and brighter (IMD is introduced also- see Norman Crowhurst; its not like this problem hasn't been known about for a very long time...). Feedback can be applied without distorting it before it can even do its job!
2) SETs have a quadratic non-linearity (Geddes hints at this in the article at the first link you provided to me above). This results in a prominent 2nd with harmonic amplitude decreasing as the order of the harmonic is increased, on a curve that Cheever likes. But if the amplifier is fully differential, it will express a cubic non-linearity, resulting in a prominent 3rd as opposed to a second (this is because even orders are cancelled, not just at the output section but throughout the circuit). As the order of the harmonic is increased, the amplitude falls off at a faster rate than seen with a quadratic non-linearity (IOW is inherently lower distortion since distortion is not compounded as much from stage to stage in the circuit). This results in smoother sound as the harmonics are more effectively masked and greater transparency since distortion obscures detail. In case there is any question about the 3rd harmonic, this is also the primary distortion component of any properly functioning analog tape machine...
What is important to note is that when the amp has a single-ended input and push-pull output is that both non-linearities are present, resulting in a more prominent 5th (see Norman Crowhurst).
IOW he didn't cover all the bases; IOW SET isn't the only answer! If the feedback signal isn't distorted, the resulting distortion will be the same as the innate distortion of the circuit but simply at a lower level.
Actually, clearly it is! If you do a search on SETs, you'll find that a very common comment about them is how 'dynamic' they are for the amount of power they have! That 'dynamic' quality is the direct result of distortion on transients and nothing else- because the speaker lacks the efficiency to allow the amp to operate at a lower level where this does not happen.
Yeah, and this is why (IMO) "voicing" is important. If you've ever listened to something like a Benchmark amplifier - really any amplifier designed primarily for absolute minimal distortion - you'll generally hear the opposite effect: dead, compressed sound. So the trick is to strike the right balance in order to.soumd natural. Of course, there are many other factors that effect the final sound but design point and resulting distortion is certainly one.
Similar threads
