Print
Category: Reader Feedback Reader Feedback
Created: 16 March 2021 16 March 2021

To Doug Schneider,

What a great review by Evan McCosham. He’s a keeper. Speaking for myself, I guess, he’s listening for the sort of things I listen for—which is one major hurdle—but he also shows a good comprehension of the product and what other consumers will look for. Good on him, and on you for not setting him up with dumb comparisons. For me, something like a $4k class-AB amp from a prestigious-but-not-unattainable manufacturer would be *exactly* right, as was the Purifi demo unit. Honestly, all of this shouldn’t be as rare as it is, but damn, that’s the truth.

Your new measurements: you guys have outdone yourselves. Just a total leap in sophistication, depth, and explanation. I guess there’s some AudioScienceReview.com issue with you guys using up to 90kHz for THD+N, but I have no idea about any of that. All I know is that I now understand way more about the amp than I ever could have hoped, or would have gained elsewhere.

Kudos, and all the best,

Tony A.
United States

I agree—Evan did a great job. And this was his first review! I also agree with you about having an appropriate equipment comparison; otherwise, what’s the use?

As for the AudioScienceReview.com “issue,” I looked through their forum and found the comments I think you’re referring to, which I’ve linked here. It mostly has to do with the bandwidth used for our amplifier measurements, which we report on our SoundStageNetwork.com site; specifically, in the chart labeled “THD+N ratio (unweighted) vs. output power at 1kHz into 4/8 ohms” that we presented on the measurements page for the NAD C 298.

When Diego Estan measured the C 298, he used a bandwidth of 10Hz to 90kHz for most of the measurements he produced. It’s the default bandwidth that he’s used for other amplifiers, but when someone on AudioScienceReview.com measured the C 298, they used a 22kHz upper limit for the measurements. (I’m not sure what the lower limit was, but that’s not important in this case.) The significance of the bandwidth has to do with the nature of class-D amplifiers, a category that the C 298 falls into.

Class-D amplifiers typically have elevated noise levels above 20kHz, which is above the audioband—in other words, beyond what we can hear. Therefore, setting the bandwidth limit to 22kHz filters that high-frequency noise out of the measurement, resulting in lower THD+N (total harmonic distortion plus noise) figures within the audioband. Extending the bandwidth to 90kHz allows that high-frequency noise to creep into the measurements, resulting in higher overall THD+N figures.

The issue, then, is what bandwidth to use. Unfortunately, there’s no easy answer. Some authorities feel that the inaudible noise above 20kHz doesn’t matter, so it shouldn’t be included in the measurements, and, therefore, a 22kHz (or similar) filter should be used. Others feel that the high-frequency noise is relevant and should be included in the measurements, hence the necessity to increase the bandwidth to capture it.

We had considered all of this before we began our measurements program. When the issue was brought up for this amp on AudioScienceReview.com, we revisited it and, as a result, will slightly revise the way we present the THD+N data for class-D amplifiers we measure in the future. We will still use the 90kHz bandwidth for the chart that shows THD+N versus output power, but we will show an additional chart that also uses the 90kHz upper limit, but applies what’s called “A-weighting” to the resulting THD+N figures. This isn’t quite the same as setting the upper-limit filter to 22kHz, which we could easily do, but does attain more or less the same result. We believe this should satisfy those people who want to see what the THD+N looks like with the high-frequency noise omitted.

The purpose of A-weighting is to account for the relative loudness perceived by the human ear, as the ear is less sensitive to low frequencies and very high frequencies. A-weighting puts the emphasis on the frequencies we are most sensitive to, rather than those close to and beyond the normal limits of the audioband, which includes the high-frequency noise above 20kHz. A-weighting is very commonly used in published audio specifications that relate to noise. In fact, in our table of what we call “primary measurements” in each measurement suite, we have already been doing this by showing both full-bandwidth and A-weighted measurements for signal-to-noise ratio, noise level, and THD+N. So we’ll simply include one more chart.

As I said before, there doesn’t appear to be a right or wrong answer for this issue in terms of the “correct” bandwidth to use, so adding more information seems to be the best solution. . . . Doug Schneider