Enrolling May 5 to May 12

The GedLee metric - a new way to measure distortion?

The GedLee metric - a new way to measure distortion?

Distortion measurements do not always tell the whole truth about what equipment sounds like. Should we pay more attention to some distortion products than others?

by David Mellor, Course Director of Audio Masterclass

Hi-Fi News a few months ago had an interesting article on the measurement of distortion and the correlation of such measurements with subjective experience.

Distortion occurs in all electronic and digital audio systems. The shape of the waveform of a signal is altered leading to the production of new frequencies from the frequencies present in the signal.

Distortion is measured by inputting a sine wave, which is the simplest signal consisting of only one frequency, into the equipment under test. If this is 1 kHz, then the harmonic distortion products generated will be 2 kHz, 3 kHz, 4 kHz... all the whole-number multiples of 1 kHz up to and beyond the limits of human hearing. The level of the distortion components is then expressed as a percentage of the whole signal. A figure of 0.1% is considered adequate, but even lower figures are desirable.

But there is a problem. It is possible for a piece of equipment to measure very well, yet still not sound entirely clean. The suspicion is that not all distortion products are equally offensive to the ear.

Listening tests conducted by a number of researchers tend to show that higher-order harmonics are more troublesome than lower order such as 2nd, 3rd etc. This implies that it would be helpful to weight the individual harmonics in the measurement according to their nuisance value.

This is not new science. As early as 1937, according to the Hi-Fi News article, the Radio Manufacturers' Association of America suggested that higher order harmonics should be progressively more heavily weighted. The BBC refined this in 1950 and again in 1961.

The problem was however that in these early days of electronics it was extremely difficult to measure the levels of the various harmonics accurately. Even today, this is not a standard facility on most test equipment.

However, Earl Geddes and Lidia Lee have suggested that not only is this essential, but that distortion products at low levels also be more heavily weighted, and have come up with the 'GedLee metric' which is basically a formula for calculating a figure from measurements that will correspond closely to the subjective experience of distortion.

This is a worthy enterprise. If there could be one figure that all manufacturers could quote and could be relied upon, it would be very much easier to make comparisons between different equipment.

However, the chances of such a common system being adopted are not good. For one thing, manufacturers whose current products do not fare well under the GedLee metric will not be inclined to use it. For another, the GedLee metric is not the 'god equation' of high fidelity. There could well be better ways of measuring the subjective effect of distortion.

Perhaps the ultimate solution would be to publish tables of harmonic spectra, then users could calculate their own figures based on whatever metric they think is best for their purposes.

By David Mellor, Course Director of Audio Masterclass
Tuesday November 21, 2006

Readers' comments on this article...

Ian Bell, Northrepps, UK
Friday November 24, 2006

Measuring harmonic distortion is in principle straightforward. You feed in a sine wave to the amplifier under test. At its output you connect a very sharp notch filter to remove the original sine wave. Anything left is distortion.

In practice it is not so easy. You need to start with a very low distortion sine wave which is itself a non-trivial design exercise. Then you need a very sharp notch filter. If you want to measure to 0.1% distortion then this filter must attenuate the original sine wav by at least 60dB, for 0.01% it is at least 80dB - whilst leaving any harmonics unaffected - another non-trivial design exercise.

Lastly, what is left is not only distortion but noise from the amp. A simple meter will read the combined result of borh. If your noise is at -80dBu and your distortion is at 0.01% they will be indistinguishable by this method. So in a more sophisticated set up you would do an FFT on the output so you could pick out and measure individual harmonics.
Norris Turney, Southfield, US
Thursday November 23, 2006

Since we're talking about Distortion, I really would like to know just how do you measure distortion

in equpment, and what type of meter would you use?, and what type of scale?.
Ian Bell, Northrepps, UK
Wednesday November 22, 2006

My apologies if I offended, that was not my intention. In Geddes presentation (available at his web site) he says THD and IM distortion are both inadequate methods of relating measurements to perceived distortion. His central premise is that it is the transfer function of the amplifier that is important and he proposes a metric based upon this. He then conducted some listening tests using various transfer functions which demonstrated that perceived distortion was poorly correlated with THD and IM distortion but correlated very strongly with his metric. However, he does not say how to measure this metric or to derive it from other measurements like THD or IM distortion.

Ian
David Mellor, Record-Producer.com
Wednesday November 22, 2006

Now now, settle down please.

The GedLee metric assesses harmonic distortion, not intermodulation distortion.

The article isn't about intermodulation distortion, so intermodulation distortion isn't mentioned.

I don't mind negative comments on articles, or corrections where appropriate. But to say an article is 'half assed' on the basis of something it doesn't even pretend to cover seems a little unreasonable.
Ian Bell, Northrepps, UK
Wednesday November 22, 2006

This article comments only on the distortion caused by a single frequency. I don't see anywhere where it says that intermodulation distortion is important. Sounds like Anton C. is bitter that someone cares enough about this site point out its defects in the hope it might get better.

Ian
Anton C., Germany
Wednesday November 22, 2006

The article is a reasonable comment on the subjective effect of harmonic distortion. I don't see anywhere where it says that intermodulation distortion isn't important too. Sounds like Ian Bell is bitter about something and wants to kill this site that I always found interesting and entertaining.
Ian Bell, Northrepps, UK
Tuesday November 21, 2006

I was refering to the article here, not the one by Geddes et al.

Ian
Anton C., Germany
Tuesday November 21, 2006

Lidia Lee and Earl Geddes have presented their work to the Audio Engineering Society - http://www.aes.org/events/115/papers/SessionF.cfm

Maybe 'half assed' Ian Bell is more clever than they.
Ian Bell, Northrepps, UK
Tuesday November 21, 2006

Oh dear, another half assed technical article. Whatever gives people the idea that measuring harmonic distortion is the only test of sound quality available. For example, for music, intermodulation distortion is far more audible but is it ever specified or measured - no. Back in the 50s and 60s is was a STANDARD measure for prefessional tape recorder and amplifier performance. And then there is square wave performance - a very simply highly revelaing test but rarely mentioned these days.

The problem is not that equipment can measure well yet not sound entirely clean - the problem is people just don't bother to make a range of performance measurements.

Ian