It's how you get there

January 21, 2017
 by Paul McGowan

If the engineer's goal is to have as little distortion as possible, then the amplifier or loudspeaker with the lowest number wins. Right?

Wrong. Just ask the receiver manufacturers who boast of vanishingly low THD but aren't worth the metal they're made from when it comes to high-performance audio.

Designing audio equipment that sounds good is an art. A balancing act.

The tools we have at our disposal to affect THD are many, but not all of them are something we want to use.

For example, negative feedback. Judiciously applied it's good. Overused it makes for hard and bright sound.

An input stage without any negative feedback might produce a relatively high THD level of 0.1% - but sound better than one with lots of feedback measuring 0.001%

Today's takeaway: it isn't the measurement that matters. It's how you achieved it.

Subscribe to Paul's Posts

14 comments on “It's how you get there”

  1. An oscilloscope's task is among others to amplify the incoming signal. I wonder what are the quality parameters of a high precision oscilloscope? Slew rate? Has anybody measured and analyzed the colorations and distortions created within the recording and mastering process? Colorations and distortions that now has to be masked by the inevitable colorations and distortions added by the stereo equipment? Now voicing comes into the game! Based on a specific loudspeaker (or headphone?) having it's own coloration and distortion in the specific listening environment! Sounds most mysterious!

  2. It's not mysterious. It's the best a man can do.
    Accept that perfection does not exist on this planet.
    Anyone who wants the "perfect" measurement for the "perfect" soundsystem, should look for another hobby/living.
    Or be happy with what PS Audio and others come up with.
    Or be disappointed for the rest of your life.

  3. Human hearing mechanism is not very sensitive to low order harmonic distortion, 5% of second order is nearly inaudible so THD is a bad optimization parameter. We are supremely sensitive to transient, intermodulation and high order harmonic distortions, so these should be the specs we read and pay for. I barely remember seeing a TIM spec, and never saw intermodulation graphs on speakers. I had a report of 110% Doppler distortion measurements on a popular consumer brand (rhymes with "nose") from an AES Fellow. We should also look at distortion of MilliWatt signals because that is most of audio.

    My big peeves are temporal distortion, transient distortion, spatial distortion and anharmonic resonances. This is where the audio paradigm performs so poorly the marketing departments would never let the specs out if the engineers bothered measuring them. In fact, most real transients are filtered out of recordings because even studio monitors do a poor job on them!

  4. It seems that a further complication is that a system is essentially a sum of these distortions, with heavier weight on components closer to the source, as they get "amplified" by every component that follows them. Over time, I have observed that distortion ratings tend to have a bit more merit for digital sources vs. an amp. While that is not always true, distortion and jitter numbers are what I give a bit more merit for a DAC or CD player. I have noticed both those 2 numbers continually get better as digital playback has improved over the years.

  5. I'm told by people I trust that anything can be measured. The problem lies in choosing what to measure and how you measure it.

    Where it turns into bull-pucky is when somebody brags about distortion measurements or amateur-grade double blind listening tests "proving perfection." For example when there is a delay in the onset of distortion a traditional harmonic distortion meter may simply not show it. We can be very sensitive to timing and filter artifacts but this varies a lot among individuals in addition to being something that is learned over time.

    I've learned over the years to believe that people actually hear the things they claim to hear but there are often crazy theories about the causes. We recognize something changed but it could be a tenth of a dB. in level, a slightly different distortion spectrum or even a tiny speed change.

    1. I always learn something reading these posts. Today I learned bull-pucky, which was a new one for me. I shall have to find a way to use this in real life this week! Thanks! 😀

  6. I can't help , but wonder, if there are things that we hear that have not been identified or isolated as a measurable parameter that can be use to further improve sound quality. There must be more to sound than THD/IMD , slew rates, jitter, damping factor, etc.

    1. Yep - if no one has realized that a particular thing is an issue or has thought of a test for it, it is outside of the current things to test, which is a subset of the possible.

      e.g. The Beatles' engineer hearing/experiencing/being bothered by what turned out to be a hump in FR beyond 20 kHz, compelling the equipment designer (Rupert Neve) to build test equipment with FR to 50 kHz, whereupon he then found the hump the engineer heard in his console (desk) design up around 30kHz.

    2. There are numerous things that have been identified but figuring out exactly what factor is causing a change can be a very costly proposition. Adding to the confusion is the fact that no two things ever actually measure the same.

    3. It's the height of human arrogance to believe that we have and can "measure anything". Even that, which we don't know exists.

      I am reminded of a famous quote from eminent physicist Albert Michelson (Michelson-Morley experiment) in 1894, at the dedication of the University of Chicago's Ryerson Physical Laboratory:

      "While it is never safe to affirm that the future of Physical Science has no marvels in store even more astonishing than those of the past, it seems probable that most of the grand underlying principles have been firmly established and that further advances are to be sought chiefly in the rigorous application of these principles to all the phenomena which come under our notice. It is here that the science of measurement shows its importance — where quantitative work is more to be desired than qualitative work. An eminent physicist remarked that the future truths of physical science are to be looked for in the sixth place of decimals."

      Indeed, the idea there wasn't much more to measure or discover in physics (this before Einstein came onto the scene) simply because many of his era were out of ideas - figured they had it all measured and locked up - is laughable now.

      Just like the idea we can measure pretty much anything today.

      Might have to work that into tomorrow's post.

  7. Does how you get there matter? Yes and no. You can get across a river swimming the crocodile infested waters and risk get eaten alive, you can take a boat and risk capsizing, or you can walk across the bridge. Where are you now, where do you want to get to, and what is your plan for a path getting there. If you don't know you are wandering aimlessly. You're lost in the forest and there is no yellow brick road to get you to Oz. One way looks like any other. You see a sunny clearing? Does that mean you're near the edge of the forest or is that where the wicked witch is waiting for Hansel and Gretel, or where the Jabberwock with jaws that bite and claws that catch has his lair?

    What do engineers do? (I mean the real ones, not the wanabees?) They study where they are to get their bearings, they study where they want to be, and they study all possible paths to get there (there are often more than one) and then they pick out the best one and follow it. Technically the path is called a transfer function, a mathematical path or plan. What transfer function do you want? If you are designing a sound systems for a 1980s style Discotheque you want one that produces earsplitting loudness and heart pounding bass. You want sound so loud people can't think and any sane individual will do anything possible to escape it. I had a friend who turned down a request to design a sound system for a famous NYC Disco that would produce 150 db of sound. Sound can be extremely dangerous. At 220 DB it is lethal to humans.

    What are some of the considerations engineers must take into account when making choices. Reliability, cost, adaptability, adjustability. Often there are tradeoffs. (I do this with other kinds of systems as my day job folks. I face the same problems. Once a simple 75 KVA transformer that can be found in just about any commercial or industrial building in America was a no brainer. All there was was brand. Now there are at least 30 separate choices for generic types and they range from very expensive to rediculous but given their differences, the wrong choice can spell failure on one hand or a waste of a lot of money on the other, in fact enough to buy a pair of BHK monoblocks.)

    A word about negative feedback since this always seems like a mystery to tyros. Negative feedback is a method of nulling out distortion. The Price you pay for it is gain. By adding the difference between input and output back to the input out of phase you predistort the input signal in the opposite sense so that at the output you get the signal you wanted in the first place. Too little negative feedback and you have not optimized the design. Too much and you are adding distortion back in, you've gone beyond the optimal null point. Is it simple? No. Is it powerful? Yes, very powerful. Like all powerful tools results depends on who is using it You can use a chain saw to cut down trees in a forest to build a log cabin or you can cut your hand off and bleed to death. Tyros should not go near powerful tools. They get hurt that way.

    Here's the equation in a nutshell G'=G/[1+GH] Looks simple doesn't it? well it's not. Each of those letters is a differential equation (calculus) usually expressed in the Fourier transform format. G' is the overall gain of the system. G is the forward Gain. H is the reverse gain. GH is the loop gain. Here's an interesting problem, if H does not have the same frequency response as G, then while you are nulling out distortion at some frequencies, you are adding to it at others. If the net gain G has a gain of 1 or more when the phase is 180 degrees you have built an oscillator. Get close to 1 and you have built a resonant circuit that will oscillate with a signal at the system resonant frequency that will decay over time. Add a load and if is associated with the feedback circuit you may have just altered H....or G...or both. An ultrasonic resonance that is inadvertently created by a high capacitance load such as from some audiophile speaker wires can drive the system to full output and blow it up. The text I used was written to two electrical engineers who happened to be Generals in the US Air Force. It was intended as a text for seniors and graduate students in electrical engineering and it is hard, very hard. There's a lot of math in it.

    It should be noted that the transfer function that must be considered is the acoustic transfer function between the performer and someone in an audience if the system is to be classified as high fidelity. To achieve that you must first understand that acoustic transfer function (of sound fields) and duplicate it using various recordings, types of recordings, all made different ways, and the acoustics of your listening room. If you don't do that, you don't have a prayer of achieving high fidelity no matter how much you like the results. This is why the more you understand about the problem, the more feeble, even laughable the efforts of those in this industry trying to solve it are. They haven't even got a clue to what the problem is actually about let alone have any chance to solve it. They're always eaten by the crocodile or the Jabberwock.

  8. Here's another way to look at harmonic distortion. Suppose a gain stage say in a recording studio has 0.25% THD, then its output signal is 99.75% non THD. Now suppose the next stage also has 0.25 THD. Then the output of two stages has .9975x.9975 non THD = 99.5% non THD. Each stage the signal passes through reduces the non THD from the original signal by whatever is left after the THD it contributed is subtracted from the overall signal. The more stages and the more THD per stage, the more distorted it gets. This is for analog only. So what threshold do engineers want to design to for an overall system including all recording and playback. They want it to end up with THD below audibility. So where does that lie? I've read it's about 6% THD. This is why even the most miserable tube amplifiers which can add at least 0.5% and are typically 1% to 2% THD still have inaudible THD. So if 94% or more of the original signal is still there you're OK. 20% is 1 db and since dbs are a logarithmic scale 6% is a small fraction of a db. (I know, you golden eared audiophiles can hear a gnat's eyelash when it blinks.)

    One of the most important performance specification for a woofer is THD. This is because woofers have a tendency to break up and vibrate in harmonic modes. Acoustic suspension speakers are less prone to this for several reasons but one of them is that the restoring force on the cone is applied uniformly over its entire surface. This lack of difference between restoring force any two points means there will be no shearing or twisting force applied due to the suspension. By contrast, differences in restoring force around the perimeter of the inner and outer suspensions of non acoustic suspension drivers and difference between corresponding points on the inner and outer suspensions cause the cone to breakup. Usually acoustic suspension woofer cones are heavier and therefore stronger than others because the free air resonance needs to be low. Typically an AR 12" woofer has a free air resonance of 16 to 19 hz. Early variants had about 5% THD at 30 hz. More recent variants are around 1%. This means that servo feedback is entirely unnecessary because almost all of the distortion is linear distortion that can be corrected with optimal enclosure design and equalization. When other types are measured, they can have as much as from 40% to 100% THD driven at 30 hz. This is one reason why they can't produce deep bass.

Leave a Reply

Stop by for a tour:
Mon-Fri, 8:30am-5pm MST

4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

Stop by for a tour:
4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram