Measuring amplifiers

“The best amplifier is a wire with gain”. It’s a saying from Steward Hegeman of Harman. It is an absolute truth. Unfortunately, this wire with gain does not exist. An amplifier unfortunately needs a bit more. And because of that, things can go wrong.

Various flavours

We won’t go into the various amplifier classes here (class A, (A)B, D and in the niche: C, G and T). But it is useful to know what the type of amplifier is before you start measuring.

A class D amplifier sometimes requires a slightly different set-up than a class A or (A)B one. Class D works with a high switching frequency that in certain cases can influence the measurement namely. By putting a special (kind of low-pass) filter in between, you remove that noise. We usually measure without it (after all, you also just connect the speaker like this), but we have a filter if it turns out that the measurement does not go well.

The measurements

We measure a few things in amplifiers:

• Noise floor
• Distortion (THD+N) versus frequency and amplitude.
  • This includes SINAD (Signal Noise and Distortion).
• Distortion versus power
• Maximum power
• Crosstalk (channel separation)
• Frequency response
• Phase response
• Residual

Example: MSB 202 power amplifier

This is an example of an (ultra) high-end power amplifier. We see a noise floor (at a 1 watt – 8 Ohm measurement) of about -130 dB which is incredibly good. We also see before the 1 kHz input (test signal) some harmonics from the 50 Hz mains input. Those bumps are below -100 dB. We do see these harmonics of 50 Hz quite often. Sometimes it is a grounding problem in the measurement setup. We check that by grounding the device on the Prism then too. However, sometimes this is not the problem and it is simply in the design. It’s not abnormal; we see it quite often with class AB amplifier with a toroidal transformer.

Crosstalk A to B is very good with almost 130dB separation. This will manifest itself in a very tight stereo image. Many amplifiers barely reach 100 dB. So this is exceptionally good.

The phase measurement shows that up to 20 kHz there is almost no phase shift (max 14 degrees.) Class D does much less well here because of the filter that is needed. Every filter causes phase shifts. That includes a filter in an amplifier.

Then the residue measurement. This is a wacky one. In fact, this measurement shows whether crossover distortion is present. Ideally, a sine wave comes out. Then the residue is distortion-free. In class D and AB amplifiers with crossover distortion, we will see a distorted sine wave or other ‘junk’. This will manifest itself in a somewhat rough character. You will see this further on.

Then Linearity. This shows how accurately the amplifier follows the input. Ideally, we see a diagonal line from bottom to top. However, an amplifier cannot track below the noise floor. After all: there, the noise is stronger than the input signal, so the amplifier sees no input and cannot ‘follow’ that signal. So this is also a measurement where the noise floor is reflected!

The linearity measurement is also an important measurement to see whether a device is operating neutrally and cleanly. And how ‘deep’ into the soft passages a device can follow well!

Distortion is fairly self-explanatory. We measure this with a 1-watt sweep (4 and/or 8 Ohm loaded). You can then see the distortion from 5 Hz – 20 kHz or higher, depending on what we set. Usually we measure up to 20 kHz.

What a good score in the distortion measurement is, is a bit dependent on the technology. Class D amps score almost always extremely well in terms of distortion. They sometimes go as low as 0.003%. A ‘normal’ class A and A/B amplifier has a bit more distortion. This is inherent in the technology and the amount of feedback a manufacturer applies.

To illustrate, around 0.01% is extremely good with class AB. The bulk is at the bottom around 0.01 / 0.03% and peaks to around 0.05%. Peaks around 0.05% are partly due to the low feedback applied by some manufacturers. Less feedback results in higher distortion. But mostly the amp does play more smooth and playful.

Also know that the type of distortion plays a role. Which harmonic is dominant? That partly determines the colourfulness of an amplifier and it can make the distortion not noticeable at all (we don’t find 2nd harmonics annoying).

You see: it’s not as black and white as we all often think (and hope!).

Finally, maximum power. Here it’s a bit wild, wild, west…. How manufacturers specify this is completely unclear. We measure at the ‘cross point’; the kink in the graph. After all, that’s where distortion shoots up. But a lot of manufacturers stick to <1% THD. And so that may already be in the upward trace.

We measure on 8 and 4 Ohm load. For this, we use non-inductive dummy loads that can take 250 watts continuously. We have them heavily cooled, so peaks above 1 KW are not a problem: in fact, they do not get hot at short loads.

Class D – Hypex Nilai

Nilai phase

For comparison, above, a Hypex Nilai. You can see how unprecedentedly clean the Nilai measures. At the bottom no 50 Hz harmonics at all (switch mode power supply) and also above it (after the 1 KHz test tone) almost no ‘grass’ as we call it. (short spikes in the spectrum).

The distortion of the Hypex Nilai is unprecedentedly low. Almost a straight line around 0.02%. Very impressive, of course!

In the residue measurement, you can see what we mean by the distorted sine wave. By the way, this is still very good for a Class D! We have certainly measured worse models.

What few people know or realise is that a Class D amplifier has a limited bandwidth. At around 50 kHz, Nilai offers a lot of bandwith for Class D. The bulk of class D already drops off from 10 KHz with -3dB at 20 kHz. This is due to the filters they have to use.

Finally, the phase. With a Class D, this measurement can be directly linked to bandwidth. After all: the bandwidth is limited by the filter. And the filter affects… phase. But even here, the Nilai still does just fine. However, it is very clear that there is more phase shift  than with a class AB! The question is how this is audible in practice. We need to do more research on that.

Concluding

We hope you have now got a little more grip on the amplifier measurements. We realise that it is not always easy to make sense of the graphs. Especially as ‘noise’ is not always the same. And not everyone is at home in dBs, Hertzes, Ohms and volts. If you have any questions, feel free to ask via e-mail or the forum!