What matters

January 22, 2017
 by Paul McGowan

Here's the thing about THD distortion. It mostly doesn't matter. Even relatively high levels of THD are not recognized by the ear.

The impacts of what amp designers do to lower it are far more noticeable than the added harmonics themselves.

And loudspeakers make orders of magnitudes more THD than amplifiers.

Factoid: the next time you see higher levels of THD listed on an equipment spec be attracted rather than repulsed. Higher levels of measured THD often signal a better sounding design than the opposite. (Not because they are higher, but because the design is likely more musical).

Measurements that do matter are mostly not mentioned: TIM (Transient Intermodulation Distortion), slew rate, or open loop stability are examples.

One measurement often included, but not understood, does matter. IM. Intermodulation distortion is quite noticeable to the ear and is often the red-haired stepchild, mentioned but not angst over.

Tomorrow we'll angst together.

Subscribe to Paul's Posts

11 comments on “What matters”

  1. Paul, would there ever have been a PS Audio without Terri?

    Here’s to all the women worldwide who stood up yesterday in strength,

    One voice, one instrument, united and amplified together. 🙂

    1. Because most people wouldn't know what to do with it and without a comparison or reference to anything else it's rather meaningless. And, besides, we don't have the facility to measure it even if we wanted to.

      The trick with TIM is to make sure the circuit is open loop stable - which many amps aren't, but all ours are. Once that's been achieved, TIM is not an issue.

      1. I'm not sure how you would get an amplifier to have an open loop gain that is not stable unless you deliberately or accidentally set out to build an oscillator. You might create one if you connect the wrong amplifier to the wrong load. The output could become a "tank circuit" which merely means a resonant circuit. Maybe. A spike in FR means the system has an inherent resonant frequency at which it is prone to oscillate. How well damped that peak is would determine if it will oscillate. I think you'll find that in its pole zero diagram. Check with BHK.

        1. You have too much open loop gain - and that does it. Not necessarily what you're referring to, Mark, which is an 180˚ phase shift into an oscillator, but unstable in terms of passing a reasonable signal without simply slamming to one rail or another. Maybe "unstable" was a poor choice of words on my part.

          1. Paul, a radio station has enough gain to turn a few millivolt phonograph signal fed into a 47,000 ohm preamp load into 50,000 watts of RF output broadcast over hundreds of miles. How much more gain does it take to become unstable?

            If you are talking about overloading an inadequate power supply then anything might happen. And yes I've seen that in one of the stupidest PA system designs you could imagine....engineered by one dumb cluck of an EE and Johnson Controls. It was my job to fix it. You could hardly create a worse system design that oscillated between saturation and self protection shutdown. The fire marshal was not amused.

  2. " the next time you see higher levels of THD listed on an equipment spec be attracted rather than repulsed. Higher levels of measured THD often signal a better sounding design than the opposite. (Not because they are higher, but because the design is likely more musical)."

    Hmmmm. The truth is so obvious and so opposite. Once it reaches an audible level it is horrible. Let's start with an easy one. Remember those early solid state amplifiers that sounded so bright and harsh? Remember the oscilloscope photographs showing crossover notch distortion where the two halves of the sine wave, the positive going side and negative going side didn't quite meet up? That is what harmonic distortion can sound and look like. Since that notch becomes a greater proportion of the signal at low levels, one trick to more easily hear it told to me by an old electrical engineering friend is to play the sound softly (a more efficient speaker is better for this test) and put your ear close to the speaker. This is why amplifiers began to give THD specs at rated power and below even though tube amps could have crossover notch distortion too. But it was relatively rare, tube amplifier designers had already solved that problem. BTW, that's why class AB amplifiers often come with a bias control. It is intended to minimize crossover notch created harmonic distortion by setting the optimal bias current. Optimally it should be adjusted using test equipment.

    Another example is amplifier clipping. Next time you see a power amplifier review that has measurements, look at the graph of THD versus power output. Up to rated power it's usually a low flat line near the bottom. Then the graph shoots straight up. On an oscilloscope the top of the sine wave looks flattened. The more the amplifier is overloaded the more of the curve is flattened and the higher the harmonic distortion. One culprit for this in tube amplifiers is the output transformer among several kinds of distortions it creates. Look at a BH curve. The transformer couples the primary and secondary through an alternating magnetic field in the core. Near the top and the bottom of the curve, the core starts to saturate. More current from the plate circuit and therefore more inducing magnetic field does not result in a corresponding increase in the induces field and so the curve flattens. The output current therefore doesn't increase either and the result is harmonic distortion. This is why good loudspeaker crossover networks never use iron core inductors.

    Phonograph cartridge mistracking, analog tape overload, speaker cone breakup, speaker excursion beyond where the magnetic field in the gap is uniform all cause audible harmonic distortion. When a woofer cone is overdriven and breaks up, you don't hear 30 hz, you hear 60 hz, 90hz, 120 hz. Atkinson's measurements never show you that. He assumes if he puts a 30 hz input into the speaker, that's what he's getting out.

    The loudness wars for pop CDs to be heard on a radio deliberately overload the wide dynamic range digital system to get average louder and therefore more attention getting sound when you flip through stations. The result, very high harmonic distortion. The ipad generation who listens through ear buds to ersatz music doesn't care. Played through an expensive hi fi system you'll get perfect reproduction of the distortion inherent in the recording.

    One of the worst technical papers I ever read was written by a man named Cheevers as a thesis for a masters degree in Electrical Engineering. Curl cited it as a paradigm every amplifier designer should read. Therefore I knew it had to be awful. It has more flaws than a wheel of Swiss cheese. In spite of that, he got his masters degree anyway. (Had I been on the judging board he'd have had his BEE rescinded.) It's main thesis is that audible harmonic distortion is more objectionable for a given total THD if the harmonics are higher order. He devised an exponential scale to weigh them. What if the combined total of all of them is below audibility such as .01%? That was only one of many flaws.

    Because harmonic distortion increases the output level of harmonics that would occur naturally for most musical instruments, it's like turning up the treble, creating a rising frequency response. That makes the sound bright and harsh. Is 6% really the audible threshold? I don't know. I write about TIM later, a simple concept made complicated if I feel like posting another entry later.

  3. so true. when hp encountered the Jadis electronics, he gushed over them but someone somewhere trumpeted the measured distortion. hp contacted Jadis and they took it to heart and lowered it without corrupting the sound quality. to this day, i prefer the Defy 7 on Wilson W/Ps.

    and my preference on Wilson products is tube over ss as every time ss is used in demo of Wilson loudspeakers, they reach distortion or clipping. that's just my experience.


  4. TIM, Transient Intermodulation Disortion. What does that mean? Slew rate distortion, what does that mean? Actually they mean the same thing. And had electrical engineers done their job properly all along, the concept would have been redundant and obvious. But in the 1970s Matti Otala working for Harmon Kardon made this earth shattering discovery and HK began to market the latest silver bullet. So what is it in plain English? Simple enough. When amplifier frequency response is measured, it is usually measured at one watt output typically into a 4 or 8 ohm resistor for a load. In the 1930s the FRs were all over the place and few amplifiers put out more than a few watts anyway so real differences were apparent to anyone with access to the data. Fast forward to the transistor era and that measurement gave textbook perfect results. One amplifier looked pretty much like another by this measurement. But had the measurement been made at much higher output, say nearly at full output, things would have looked very different showing real disparities. For many amplifiers high frequency response in the audible range would have fallen off drastically. This is because the amplifier couldn't deliver lots of power at high frequencies that is suddenly. So an orchestra playing loudly suddenly confronting an amplifier with a loud cymbal crash with high energy at 11 khz too where the amplifier was already putting out plenty of power at other frequencies just couldn't deliver and the sound was less than clear. The slew rate just means that the current and voltage couldn't change far enough fast enough to reproduce what the speaker and program material demanded. Back to low power output and things were fine again. Why is this called intermodulation distortion? Because a host of new frequencies are created, in fact they'd be created even if there was only one frequency input. Why is it transient? Because it only happens during the period when output levels should be changing rapidly. Does it sound bad? Yes. Is there an easy fix for it? Yes, design better amplifiers with large oversized power supplies, big transformers and lots of capacitors and design good circuit topologies. I didn't say it was cheap, I said it was relatively easy. Can other equipment suffer the same problem as power amplifiers? Yes, it can happen at low signal levels too. Typical place to find it is the output of a preamp like the Nagra working into a relatively low input impedance solid state amplifier. Those 50 ohm emitter follower and cathode follower output stages will get you again and again. I think the Nagra has a 600 ohm output impedance, can't quite remember. Solution, a preamp with a better power supply and a much lower output impedance like a push pull amplifier output stage. Are these ideas expensive? No but they require thinking and breaking with convention. This explains why the very high input impedance of the BHK amplifier results in better performance and why I endorsed using a vacuum tube input stage until a comparable transistor configuration could be found. What's the input impedance Paul, 100K? 200K? What about the competition? 10K? 5K?

  5. Paul, I'm looking forward to hearing more about the measurable things that matter and the ones that don't. Yesterday, you said "Overused {negative feedback} makes for hard and bright sound." Does that "hard and bright sound" have a name? Is it measurable?

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