|
Home » Audio » Thermionic Emissions » Why SE in SET amps?
|
|
| Re: Why SE in SET amps? [message #62841 is a reply to message #11382] |
Sun, 23 May 2010 17:54   |
Pano
Messages: 17
Registered: May 2010
|
Chancellor |
|
|
OK, I jump in and embarrass myself in my 1st post here.
I was lucky enough to be introduced to the DHT & SET "cult" in the mid 80's by none other than Mr. Jean Hiraga himself. I was as surprised as anyone at the sound of these little amps. Quite a learning experience hanging around with Hiraga and crew.
So here is what I know about "Why Triodes, why single ended, why direct heated?" Mostly it comes down to harmonic distortion and the structure of that distortion. It's the structure that is so important. A good SET amp does not give dominant even order harmonics, it give a very regular fall off of all harmonics, odd and even. This is very important.
Back in the 30's Wegel and Lane established that a regular fall-off of harmonics is the audible equivalent of no harmonics. Each successive harmonic is masked by the one above it, so it is not heard. So up to several % of THD will sound like a pure tone if the harmonic structure is right. This work was continued in the 70's by Matti Otala and others. I'll elaborate on that if anyone is interested.
The great thing about direct heated triodes in a singled ended configuration is that they can come very close to approaching the "ideal" harmonic structure. No other device can, not even other tubes. And they can do this with little change in the harmonic structure a different frequencies and power levels. Again, something no other device can do.
But what about amps with 0.0001% THD? Shouldn't they be better? By that number, they should be. But that does not tell us what the harmonic structure is. And the ear is very good at hearing those harmonics, even if they are tiny. Also, those amazing numbers are usually taken only at 1Khz and a fixed level. That is far form the whole story. The harmonic structure determines the sound of the amp.
Short end of the story. Harmonic structure relates best to what and how we hear. THD does not. It's more complicated than that, but that gets us started.
|
|
|
|
|
|
| Re: Why SE in SET amps? [message #62868 is a reply to message #62858] |
Wed, 26 May 2010 00:38 |
Thermionic
Messages: 208
Registered: May 2009
|
Master |
|
|
| Pano wrote on Sun, 23 May 2010 17:54 | Harmonic structure relates best to what and how we hear. THD does not. It's more complicated than that, but that gets us started. A good SET amp does not give dominant even order harmonics, it give a very regular fall off of all harmonics, odd and even. This is very important.
|
Welcome to ART, Pano! A great post; thank you for making this very important statement.
It's often said that "even order harmonics are good, and odd order harmonics are bad," but that stops short of telling the whole story. While it's indeed true that odd orders are more dissonant than even orders, the big picture is that the higher the overtone series of the harmonic distortion products present, the worse the amp will sound, regardless of whether the distortions are odd or even order in nature. The other facet of this is that all harmonic distortion products are just that, distortion, and represent a deviation from the original input signal.
In light of that fact, I've always maintained the position that SET amplifiers sound good in spite of their high 2nd harmonic content, not because of it as many contend. As Pano noted, it's because their distortion spectrum typically drops off like the proverbial rock past the 2nd harmonic, leaving them relatively void of high order harmonics (both even and odd). In agreement with Pano's post, I've found during my 25 years of experience with tube amplifiers that this end is best achieved by using highly linear, low-mu triodes. How you operate them is also highly critical, with load impedance and bias point being major determining factors in the all-important harmonic distortion structure.
| Matts wrote on Tue, 17 November 2009 11:38 | My guess is there are some micro-variations in the a.c. signals that are lost when a neg. feedback loop "blends" slightly different parts of the signal together, and this causes some loss of the cues that give our brains the sense of "live" music.
|
Yep, for sure. The error signal in a global NFB loop is always time-delayed and phase-shifted with respect to the main throughput signal, so that the two cannot ever "mesh together" perfectly. This is mainly caused by the reactance of the signal path capacitors and the output transformer.
Also, as Wayne stated, the error signal will always be tainted by the back-EMF from the speaker drivers, which itself is phase-shifted as it travels back through the crossover network and speaker cables. In short, it's impossible to fix something that's broken with something else that's broken even worse.
While global NFB does greatly lower the overall THD, it multiplies the remaining distortions into higher, more dissonant orders. In a sense, you distort the distortion, so to speak. My take on the matter is that if avoiding this means you must end up with an amp that produces only 2 watts per channel, then so be it, because sound that's been repaired is never as good as sound that wasn't broken in the first place. By all means, I'd much rather listen to mindblowing sound at mundane levels, than mundane sound at mindblowing levels. Besides, I'm too old to handle it loud any more...... 
Thermionic 
|
|
|
|
| Re: Why SE in SET amps? [message #62869 is a reply to message #62868] |
Wed, 26 May 2010 01:00 |
Thermionic
Messages: 208
Registered: May 2009
|
Master |
|
|
| Pano wrote on Sun, 23 May 2010 17:54 | But what about amps with 0.0001% THD? Shouldn't they be better? By that number, they should be. But that does not tell us what the harmonic structure is. And the ear is very good at hearing those harmonics, even if they are tiny. Also, those amazing numbers are usually taken only at 1Khz and a fixed level. That is far form the whole story. The harmonic structure determines the sound of the amp.
|
Indeed! So true! THD tells you absolutely nothing about how the amplifier will sound. On that note, below is an excerpt from a white paper I wrote on THD and IMD a few years back, that I'd like to share. I apologize that much of it has already been covered in this thread, but I had to include it so that what hasn't been covered will make sense.
"Besides the psychoacoustic effects of different orders of harmonic distortions, perhaps even fewer people understand how "on-paper" distortion percentages correlate to the ear's actual perception of them. If you test an amplifier on a harmonic distortion analyzer, and it tells you there is 1% THD, does that mean there's really 1% THD? The answer is yes, and no. The analyzer measures the distortion as a voltage that represents a percentage of the main signal voltage, not its actual sonic perception. Or, as I like to put it, "Test equipment ain't got ears. People do." However, some simple math can provide a relative conversion from volts to perceived sound pressure level.
The HD analyzer tells us that we have 1% THD, which is about 40dB below the fundamental. Converting -40dB to wattage would be 1/10,000 of the full power, which with a 10 watt amp cranked to the point of clipping would be 1 milliwatt worth of distortion. Now, here's where we get down to the nitty gritty.....
These wattage/percentage figures do not tell us how they will be perceived by actual human ears, because the ear does not perceive volume in a linear manner, but on a logarithmic curve. To net a doubling of perceived volume (approximately equal to 10dB) requires 10 times the power; likewise a halving of the perceived volume requires cutting the power by 10 times. So, -40dB roughly represents an actual 16:1 ratio to the ear's perception, not a 10,000:1 ratio as indicated by the voltage percentage, because the ear hears -10dB as one-half the volume, -20 one-fourth the volume, -30dB one-eighth the volume, and -40dB as one-sixteenth the volume.
OK, now let's see what 1% measured THD really sounds like to real human ears.
100% รท 16 = 6.25%
This illustrates how the ear can very easily pick out tiny percentages of high order harmonics. Consider a nasty high order harmonic that's buried -80dB down. To the ear, a 7th harmonic at -80dB sounds like 3.1% 7th order distortion, very offensive indeed! Add all the other harmonics from 3rd through 6th with it, and you've got yourself a real mess! Then, consider that when listening to music you're not dealing with a single frequency as in an industry-standard THD test, but a very complex arrangement composed of a nearly infinite number of simultaneous frequencies. Once again, psychoacoustic effects carry far more weight than measured specifications on paper!
Finally, as if things weren't already bad enough, we have a lot more than harmonic distortions to contend with. We have non-harmonic intermodulation distortions too! Any time you put two frequencies together you create two new ones, which are the sum of the two and the difference between them. These are, not surprisingly, called the quadratic sum and difference frequencies. Most of the time, the resulting intermodulations are musically unrelated to the original frequencies and therefore horribly dissonant.
Let's use a 440Hz sinewave as an example, which is the tuning standard for musical instruments. It is the open A string of the guitar, the A note above middle C on the piano, and two octaves below the open A string of the violin.
If we add an 880Hz sinewave with it (which is an A note exactly one octave higher), we'll have the original 440Hz and 880Hz frequencies, plus the sum frequency of 1.32kHz and the difference frequency of 440Hz. 1.32kHz is an E note, which is a musical fifth above the 880Hz A note and equivalent to the 3rd harmonic of the 440Hz A note. Not too terribly dissonant sounding, by any means.
Interestingly, if we superimpose that 1.32kHz E note sinewave over the 440Hz A note, we'll get:
Sum: 440Hz + 1.32kHz = 1.76kHz
Difference: 1.32kHz - 440Hz = 880Hz
Notice that 1.76kHz is exactly one octave above 880Hz, or two octaves above our original A-440 note. Therefore, we still have A notes and an E note! This illustrates how low order harmonically-related tones intermodulate to form harmonically related tones of similar nature.
Now, let's move on to a higher order, yet still harmonically related tone, intermodulate it with our A-440 note, and see what comes out. Let's use a musical third, that's in the third octave above the fundamental. This would be a C# note, which would correlate to a 5th order harmonic relative to the A-440 fundamental.
Sum: 440Hz + 2.217kHz = 2.657kHz
Difference: 2.217kHz - 440Hz = 1.777kHz
The resulting intermodulation frequencies are musically unrelated to both the original fundamental tones. The sum frequency falls between a musical E and F in pitch, and therefore that frequency will be dissonant regardless of what the original fundamentals were. The difference frequency is 17Hz sharper than the nearest actual note (an A note), so that it would sound something like two musicians playing the same piece together but with their instruments badly out of tune with each other, which makes for a rather hair-raising discordance.
From these simple examples, you can see two things become very clear when we relate this concept to harmonic and inharmonic distortions in amplifiers. One, the more frequencies that are simultaneously present in the music, the more intermodulation distortions you'll have. Two, high order harmonic distortions give way to some very dissonant intermodulations. These IM distortions are a chief reason why many amplifiers are harsh and fatiguing to listen to, and sound homogenized and congested when playing complex passages."
Thermionic
|
|
|
|
|
|
|
|
Goto Forum:
Current Time: Fri Nov 29 16:02:46 CST 2024
|
Posting Rules
Members
Search
Help
Register
Login
Get a T-Shirt!
