Distortion is "anything different from the original" and can't be summarized in a single number. Harmonic Distortion is a (particularly easy) way of doing a measurement with single tone sinewaves. HD is not a characteristic of the device, just an easy result of a method, and has not been shown to have a direct correlation with listening quality.
For recent literature, see Geddes Perception
For recent literature, see Geddes Perception
Not only distortion wise are the speakers the main problem to be solved.
Why bother with riaa correction within a fraction of a dB, if even the best speakers are at least a couple of dB off the straight line, and that only at a very small listening area? And then there is the issue of stored energy in drivers and enclosures. Plus phase shifts, for those who think those are audible. And finally, the directiviy pattern, where no perfect solution seems to exist. In short, much to do.
In my experience, it appears that all else being equal, it is the application of negative feedback that appears to cause the most problems. We have, since the dawn of sound reproduction had lots of total distortion from the playback system. Back in the '30's with horn loaded speakers dominant, the overall distortion could be contributed by both the horns and the open loop electronics. By the time of the introduction of the Williamson amplifier to the audio scene, we had about 20dB of negative feedback added. In my opinion, it appeared to be a subjective 'tradeoff' between 20dB of negative feedback and open loop operation, with some finding that 14db or 6dB or whatever actually sounded optimum. You will find that in many hi end tube amps today. Just look.
Now, why did they stop at 20 dB? It was because when you increased the feedback, you invariably got into low frequency instability in tube amps due to excessive phase shift of the coupling caps, output transformer, and power supply time constants (yes, they were important too!).
Only when we went to solid state and direct coupled inter-stages, did we get more than 20dB of feedback in a practical manner. Of course, some special tube amps, that were the original 'op amps' could generate and use more negative feedback, but this was almost never used for audio reproduction, for good reason. A coupling cap, in those days, was a relatively passive device (wax paper or Mylar) as it was pretty primitive in construction. Just a sandwich of foil and a separator (paper or mylar) wound tightly together. Subjectively, these caps were not too bad: Almost no measurable harmonic or IM distortion, bipolar, and so forth. The only real problem that went unrecognized was DA or dielectric distortion. That was ignored for audio, although it was extensively researched for cap use with analog computers (big stuff in the 50's) or sample and hold circuits (still popular today). So they used lots of interstage coupling caps. (more later)
Now, why did they stop at 20 dB? It was because when you increased the feedback, you invariably got into low frequency instability in tube amps due to excessive phase shift of the coupling caps, output transformer, and power supply time constants (yes, they were important too!).
Only when we went to solid state and direct coupled inter-stages, did we get more than 20dB of feedback in a practical manner. Of course, some special tube amps, that were the original 'op amps' could generate and use more negative feedback, but this was almost never used for audio reproduction, for good reason. A coupling cap, in those days, was a relatively passive device (wax paper or Mylar) as it was pretty primitive in construction. Just a sandwich of foil and a separator (paper or mylar) wound tightly together. Subjectively, these caps were not too bad: Almost no measurable harmonic or IM distortion, bipolar, and so forth. The only real problem that went unrecognized was DA or dielectric distortion. That was ignored for audio, although it was extensively researched for cap use with analog computers (big stuff in the 50's) or sample and hold circuits (still popular today). So they used lots of interstage coupling caps. (more later)
You are on to something, vacuphile, but we do need standards, and RIAA is one standard that we now follow more closely. However, when people tell me that 0.1dB is all important in a listening test, I have to take it with a 'grain of salt'. It is just something can can be demanded by the critics of audio differences, hoping to dismiss a specific listening test. Now, I am not going to say that you can't here 0.1dB over several octaves, but only under laboratory conditions. We just don't listen with our head in a vise, so to speak.
There's some truth in that- I can hear 0.2dB level changes in an ABX comparison, but no way I'd be able to manage that with (say) a five minute gap for switching. There's a much greater sensitivity to the issues that are demonstrably audible when you perform a direct comparison setup. And that's why, in a valid test, this must be controlled for.
I try to get my RIAA to conform tightly, but I'm much more concerned with channel to channel match, even though most cartridges aren't that close in frequency response channel to channel (or unit to unit).
I try to get my RIAA to conform tightly, but I'm much more concerned with channel to channel match, even though most cartridges aren't that close in frequency response channel to channel (or unit to unit).
Now almost everything, including 20dB feedback appeared to be acceptable through the Marantz tube era, so what happened that causes so much controversy today?
Well, back it the later '60's, we were introduced to the 'op amp' through either hybrid (usually rather expensive) or IC design.
The late Bob Widlar, a circuit design genius if there ever was one, made the uA702, uA709, and finally the LM101 (from which the uA741 was derived), that made some circuit design incredibly easy, compared to the past. This included servos and a number of other applications, BUT audio and video remained discrete, for good reason. It was in the early 70's that we tried to effectively use IC's (usually better than the uA741) for pro audio and this is where the problem 'hit the fan' so to speak. There was a good deal of economic pressure to go to IC's, BUT they failed to work as well as the discrete designs that proceeded them, and we started researching the problem.
Of course, limited slew rate was important, because of TIM, BUT we still found that some rather fast IC's still did not sound as good as simpler discrete circuits.
Standard measurements showed little difference, so it seemed to come down to how much feedback and perhaps how high the open loop bandwidth. This is where the Otala amp shines, first of its kind. And this is where we still stand today. WHY? Some here say we are imagining things, or that people like me are 'selling' things, but I disagree. There is still something to it, so use your ears to find the best solution. I will keep designing with the highest open loop bandwidth, the most linear circuits, and the lowest global feedback that I can get away with.
Well, back it the later '60's, we were introduced to the 'op amp' through either hybrid (usually rather expensive) or IC design.
The late Bob Widlar, a circuit design genius if there ever was one, made the uA702, uA709, and finally the LM101 (from which the uA741 was derived), that made some circuit design incredibly easy, compared to the past. This included servos and a number of other applications, BUT audio and video remained discrete, for good reason. It was in the early 70's that we tried to effectively use IC's (usually better than the uA741) for pro audio and this is where the problem 'hit the fan' so to speak. There was a good deal of economic pressure to go to IC's, BUT they failed to work as well as the discrete designs that proceeded them, and we started researching the problem.
Of course, limited slew rate was important, because of TIM, BUT we still found that some rather fast IC's still did not sound as good as simpler discrete circuits.
Standard measurements showed little difference, so it seemed to come down to how much feedback and perhaps how high the open loop bandwidth. This is where the Otala amp shines, first of its kind. And this is where we still stand today. WHY? Some here say we are imagining things, or that people like me are 'selling' things, but I disagree. There is still something to it, so use your ears to find the best solution. I will keep designing with the highest open loop bandwidth, the most linear circuits, and the lowest global feedback that I can get away with.
Last edited:
Hi,
Very well said, John.
That design principle is, as odd as it may seem, still much applied by designers of valve circuits. (Probably more by mere necessity than deeper knowledge in some case but still...).
I also can't help but notice that this basic principle is rooted deeper in Western-Europe than it seems to be in the US.
Cheers,
Very well said, John.
That design principle is, as odd as it may seem, still much applied by designers of valve circuits. (Probably more by mere necessity than deeper knowledge in some case but still...).
I also can't help but notice that this basic principle is rooted deeper in Western-Europe than it seems to be in the US.
Cheers,
Apparently, just having similar open loop bandwidth, as well as a similar amount of negative feedback works pretty well for solid state, as well as tubes. (what a coincidence) Of course, limited feedback demands more Class A operation and maximized linearity, no matter what you use. This is how I design solid state circuits. Still, a tube, measured alone is still more linear than any solid state device, by its physics, but in solid state we have the advantage of complementary design that reduces the distortion considerably, even without global negative feedback. What tends to fail is to use tube type circuits with solid state parts. They are just not linear enough to the job as well as tubes.
Of course, Frank. But the primary law is 'Child's law' that defines the usual delta Gm with delta I. That is more linear than square law (fets) or exponential law (bipolar transistors). It is fundamental. Secondly, input capacitance is well controlled and constant in tubes, not so with fets or bipolars.
Certainly listener fatigue is a primary example, but line voltage variation is another. On LPs, different parts of the LP have different fidelity,
even with a zero tracking error arm, because the groove velocity at the end of the LP is much lower than at the beginning.
If you try to repeat the same track to avoid this problem, the sound can still differ if the track is replayed too soon.
Last edited:
Yes, 3/2 power is much less than 2. Tubes are much more linear as a device.
Don't know if it's been posted here:
Unreal cables
And the old story of colouring CD's with green paint or marker pen to block out reflections so it sounds better is all a lie.
There was another one can't remember what it was called but a company designed a box with a $5 laser diode in the middle where you place the disc. Then you buy there special swipe cards for a hundred or so and swipe it in the card reader and it powers the diode to improve the disc the diode is reading. There was a review while ago.
The last one is somebody selling a small tub of black goo for $$ which supposedly made OPAMP's sound tubey when you put a bit of goo on the Semi.
Unreal cables
And the old story of colouring CD's with green paint or marker pen to block out reflections so it sounds better is all a lie.
There was another one can't remember what it was called but a company designed a box with a $5 laser diode in the middle where you place the disc. Then you buy there special swipe cards for a hundred or so and swipe it in the card reader and it powers the diode to improve the disc the diode is reading. There was a review while ago.
The last one is somebody selling a small tub of black goo for $$ which supposedly made OPAMP's sound tubey when you put a bit of goo on the Semi.
Depends on the configuration, no?
As a simple example, what's lower distortion (assuming a voltage source drive), a common cathode tube or a common emitter transistor? Now, how about a common anode tube or a common collector transistor, same circuit values as the first example?
- Status
- Not open for further replies.